{"skip":"Skip to content","position.title":"The agent-native operating layer you own","position.line":"One workspace for your team, priced on usage and outcomes instead of seats, running on models and machines you choose, and still working when a provider fails or raises its prices.","nav.products":"Products","nav.compare":"Compare","nav.pricing":"Pricing","nav.docs":"Docs","nav.site":"Site","nav.account":"Account","nav.dashboard":"Open dashboard","nav.signIn":"Sign in","nav.signUp":"Sign up","nav.openMenu":"Open menu","nav.closeMenu":"Close menu","nav.menu":"Menu","nav.group.workspace":"Workspace","nav.group.platform":"Platform","nav.group.solutions":"Solutions","footer.solutions":"Solutions","footer.workspace":"Workspace","footer.platform":"Platform","footer.contact":"Contact","footer.everyProduct":"Every product","footer.quickstart":"Quickstart","footer.downloadDaemon":"Download the daemon","footer.cliOverview":"Overview","footer.cliDownload":"Download","footer.releases":"Releases on GitHub","footer.email":"Email us","footer.enterprise":"Enterprise","footer.hostWaitlist":"Host waitlist","footer.selfHosting":"Self-hosting","footer.status":"Status","footer.privacy":"Privacy","footer.terms":"Terms","footer.tagline":"{title}: one workspace for people and agents, priced per organisation and by usage, on models and machines you choose.","footer.license":"MIT licensed. The venom CLI is free.","footer.legalLabel":"Legal and status","footer.noTrackers":"This site loads no analytics, trackers or third-party scripts.","solution.workspace.name":"Replace your SaaS stack","solution.workspace.short":"One workspace, no seats","solution.workspace.who":"Teams paying for ten tools","solution.workspace.sentence":"Notes, issues, calendars, deals, support and chat in one workspace that agents can run, on one fee for the organisation.","solution.sovereign.name":"Keep your data yours","solution.sovereign.short":"Self-host, your cloud or air-gapped","solution.sovereign.who":"Teams that answer for their data","solution.sovereign.sentence":"Run Venom in your own cloud account or on your own servers, network unplugged if you need it.","solution.mesh.name":"Run your own model mesh","solution.mesh.short":"Your GPUs and any provider, one endpoint","solution.mesh.who":"Teams running their own models","solution.mesh.sentence":"Put your GPUs and the providers you choose behind one endpoint, with logs, budgets and a fallback for every route.","solution.host.name":"Rent out spare GPUs","solution.host.short":"Serve the public mesh","solution.host.who":"People with idle GPUs","solution.host.sentence":"Run the daemon on a spare GPU box and serve requests from organisations that allow public nodes.","product.notes.short":"Notes and a wiki, linked together","product.notes.line":"Your team's shared notes and wiki, linked together.","product.tasks.short":"Issues, projects and cycles","product.tasks.line":"Track the work your team is doing, with people and AI agents.","product.chat.short":"Team chat with agents as members","product.chat.line":"Team chat where agents are members: channels, DMs and threads, with the rest of Venom posting into it.","product.calendar.short":"Booking pages on your real availability","product.calendar.line":"Every calendar you have in one week, and booking pages that know when you're really free.","product.people.short":"Everyone you know and how you met","product.people.line":"Remember everyone you meet: how you met, what they care about, and when you last spoke.","product.funnel.short":"Deals, pipeline and forecast","product.funnel.line":"Move deals through your pipeline, with the people involved and a forecast you can trust.","product.outreach.short":"Follow-ups that stop when they reply","product.outreach.line":"Follow-up sequences written from what Venom knows about each person, sent by email or WhatsApp, and stopped when they reply.","product.support.short":"A shared inbox for customer email","product.support.line":"One shared inbox for customer email and form requests, with replies drafted from your notes.","product.forms.short":"Forms with logic, shared anywhere","product.forms.line":"Build forms with logic, share them anywhere, and send the answers where your team works.","product.incidents.short":"Errors, alerts and on-call","product.incidents.line":"Catch your apps' errors, group them by cause, and let agents triage them with the flag and the release in hand.","product.status.short":"Uptime checks and a public status page","product.status.line":"A public status page fed by your own uptime checks, with maintenance windows, subscribers and incident updates.","product.comms.short":"SMS, WhatsApp and calls on your account","product.comms.line":"SMS, WhatsApp, RCS, Telegram and calls through your own provider account, with opt-outs and quiet hours.","product.line.short":"An AI agent that answers your number","product.line.line":"An AI voice agent that answers live calls on your number, with the knowledge, tools and voice you pick.","product.flags.short":"Feature flags and rollouts","product.flags.line":"Ship features behind flags, roll them out by percentage, and let agents change them safely.","product.experiments.short":"A/B tests on your flags","product.experiments.line":"Test a change on real users, read the result at a glance, and let agents run the experiment.","product.agents.short":"Chat with agents on your repos and computers","product.agents.line":"Chat with coding agents on your own repos and with agents on your computers, from any browser or phone.","product.cloud.short":"The workspace and the endpoint, run by us","product.cloud.line":"Sign up and get the workspace, the console and one OpenAI-compatible endpoint, with nothing to install.","product.mesh.short":"One endpoint over your GPUs and providers","product.mesh.line":"One OpenAI-compatible endpoint in front of vLLM, SGLang, llama.cpp, Ollama and the cloud providers you allow, with fallback and prompt-cache routing.","product.daemon.short":"The orchestrator on your own machine","product.daemon.line":"The orchestrator and its dashboard on a desktop or GPU box without Docker, dialling out to Venom Cloud so your GPUs serve your team.","product.cli.short":"A coding agent for your terminal","product.cli.line":"A coding agent for your terminal under permission rules you set, on a harness library you can embed in your own app.","product.clients.short":"VS Code, Zed, Chrome and Claude Code","product.clients.line":"VS Code, Zed over ACP, the Chrome extension and the Claude Code plugin, all talking to your own venom.","home.meta.title":"Venom: the agent-native operating layer you own","home.hero.title":"{title}.","home.hero.openCloud":"Open Venom Cloud","home.startFree":"Start free","home.hero.from":"From {fee} a month for the whole organisation. Start free, no card needed.","home.hero.doors":"Four ways in. Pick the one that sounds like you.","home.products.title":"Everything a team works in, and what runs it","home.products.lead":"Add-ons are in preview and included until they have a price.","home.products.workspace.title":"The workspace","home.products.workspace.lead":"What your team works in. Each product has its own MCP server, so an agent connects only the ones it needs.","home.products.platform.title":"The platform","home.products.platform.lead":"What runs it: the hosted cloud, the mesh and its router, the daemon on your machines and the venom CLI.","home.products.more":"The full list, with MCP endpoints and add-on status, is on the","home.products.moreLink":"Products page of the docs","home.products.end":".","home.sov.title":"Your data stays where you put it","home.sov.lead":"The same software runs in our cloud, in yours, or on your own servers. Nothing leaves without a key and a routing rule you set.","home.sov.link":"How sovereignty works","home.sov.run.title":"Run it where you choose","home.sov.run.body":"Venom Cloud, your own AWS account from one CloudFormation stack, a Kubernetes cluster, or a Linux server from one command.","home.sov.airgap.title":"Air-gapped if you need it","home.sov.airgap.body":"Build the bundle on a connected machine, check it by checksum and signature, and install it where there is no network.","home.sov.telemetry.title":"No telemetry, and tested that way","home.sov.telemetry.body":"No update checks or analytics, and the test suites run with the network cut so a new outbound call fails the build.","home.sov.record.title":"Every action on the record","home.sov.record.body":"A map of every host a session reached, a signed log of every command and file write, and keys sealed in your vault.","home.mesh.title":"Your models and any provider, behind one endpoint","home.mesh.lead":"The workspace's agents and your own apps call one OpenAI-compatible URL. Behind it sit your GPUs and the providers you choose, and when one fails, the next target in the route answers.","home.mesh.own":"Run your own model mesh","home.mesh.host":"Or put your idle GPUs on the public mesh","home.faq.title":"Questions people ask first","home.faq.what.q":"What is Venom, in one line?","home.faq.what.a":"A workspace for your team (notes, tasks, chat, calendar, CRM, support, forms and incidents) that people and agents both work in, on a model layer you control, priced per organisation and by usage, never by the seat.","home.faq.oss.q":"What is open source?","home.faq.oss.before":"The venom CLI is free and MIT licensed, and it builds on pi, which is MIT too. ","home.faq.oss.link":"More on the CLI","home.faq.oss.after":". The workspace and the control plane are the product you run or buy: their source repository is private today, and Community, the self-hosted tier, has no licence fee.","home.faq.see.q":"What can Venom see?","home.faq.see.a":"On Venom Cloud, the workspace data you put in it and your request logs, because they live in its database. Provider keys you give a daemon never leave it. On your own cloud or servers, Venom sees nothing unless you connect a daemon to Venom Cloud yourself.","home.faq.leave.q":"How do I leave?","home.faq.leave.a":"On your own cloud or servers there is nothing to leave: it keeps running, and your data is in your database. On Venom Cloud, the console and the admin API export your data, and any OpenAI client only needs a different base URL.","home.faq.host.q":"Can I get paid for hosting GPUs?","home.faq.host.a":"Not yet. The cloud already counts what a public node earns at its listed price, but payouts to hosts are a later phase. Join the host waitlist on the hosting page and we will tell you when they open.","home.cta.title":"Start with one team and one tool","home.cta.body":"Sign up free, bring one tool's data over, and invite the people and agents who use it. Running a larger organisation?","home.cta.talk":"Talk to us","home.cta.end":".","home.teaser.title":"Your team grows. Your software bill doesn't have to.","home.teaser.body":"From {fee} a month for the organisation, with unlimited collaborators and agents. On top you pay for the inference, messages and storage Venom supplies; your own keys and GPUs add nothing. Start free with {free}, no card needed.","home.teaser.cta":"See pricing","plan.mostTeams":"Most teams","plan.contact":"Contact us","plan.unlimited":"Unlimited collaborators and agents","plan.forever":"forever","plan.billedAnnually":"a month, billed annually (or {monthly} billed monthly)","plan.people":"Built for up to {count} people","plan.workspacesAll":"As many workspaces as you need","plan.workspaceOne":"One workspace","plan.workspaces":"Up to {count} workspaces","plan.auditContract":"Audit log retention agreed in the contract","plan.auditDays":"Audit log kept {days} days","plan.limit.daemons":"Up to {count, plural, one {# daemon} other {# daemons}}","plan.limit.daemonsAll":"Unlimited daemons","plan.limit.keys":"Up to {count, plural, one {# provider key} other {# provider keys}} in the vault","plan.limit.keysAll":"Unlimited provider keys in the vault","plan.limit.requests":"{count} requests a month","plan.limit.requestsAll":"Unlimited requests","plan.freeSummary":"{daemons, plural, one {# daemon} other {# daemons}} and {requests} requests a month","plan.community.blurb":"One developer, or a small team trying it with their own keys and GPUs.","plan.community.extra1":"The CLI, orchestrator and cloud, self-hosted at any size","plan.community.extra2":"Your own models, daemons and provider keys","plan.community.extra3":"Community support, no SLA","plan.team.blurb":"Every product, for a team that wants one place to work with its agents.","plan.team.extra1":"A managed workspace with every product","plan.team.extra2":"LLM gateway, logs and agents","plan.team.extra3":"Venom's provider keys, billed by usage","plan.team.extra4":"Standard support","plan.business.blurb":"Several workspaces and a year of audit history, for a company with more than one team.","plan.business.extra1":"Everything in Team","plan.business.extra2":"Central permissions across workspaces","plan.business.extra3":"Higher capacity","plan.business.extra4":"Priority support","plan.enterprise.blurb":"Procurement, security review, or a deployment of your own.","plan.enterprise.extra1":"Everything in Business","plan.enterprise.extra2":"A dedicated deployment, BYOC or private","plan.enterprise.extra3":"A contractual SLA and a security review","plan.enterprise.extra4":"Onboarding and custom integrations","deploy.included":"Included","deploy.from":"from {price}","deploy.cloud.name":"Venom Cloud","deploy.cloud.unit":"in your plan","deploy.cloud.where":"Your organisation's workspace data lives in Venom Cloud.","deploy.cloud.p1":"Daemons keep your models, and your provider keys if you choose, on your own hardware","deploy.cloud.p2":"Nothing to run or patch","deploy.cloud.p3":"You pay for the plan and the usage you meter, nothing else","deploy.byoc.name":"BYOC, managed","deploy.byoc.unit":"a month, on top of the plan","deploy.byoc.where":"Runs in your own AWS, GCP or Azure account. Your data stays in that account.","deploy.byoc.p1":"Venom operates it: upgrades, backups, monitoring","deploy.byoc.p2":"You pay your cloud provider's bill directly","deploy.byoc.p3":"Contractual SLA on Enterprise","deploy.private.name":"Private or on-prem, managed","deploy.private.unit":"a month, on top of the plan","deploy.private.where":"Runs in your datacenter or a private network. Your data stays inside it.","deploy.private.p1":"Venom operates it with the access you grant","deploy.private.p2":"You pay for your own hardware and network","rate.notBilled":"Not billed yet","rate.each":"{price} each","rate.own.label":"Your own keys and daemons","rate.own.price":"$0 markup","rate.own.note":"Inference on your provider keys or your hardware is never marked up.","rate.supplied.label":"Inference and compute Venom supplies","rate.supplied.plus":"Provider cost + {pct}%","rate.supplied.cost":"At provider cost","rate.supplied.note":"Only when the cloud serves on Venom's provider keys or the managed public mesh.","rate.message.label":"Comms messages","rate.message.price":"{price} a message","rate.message.note":"Through Venom's provider account. On your own Sinch or gateway you pay them, not us.","rate.call.label":"Call minutes, Venom Line included","rate.call.price":"{price} a minute","rate.call.note":"Same rule: billed only on Venom's provider account.","rate.storage.label":"Storage","rate.storage.price":"{price} a GB-month","rate.storage.note":"Files in Notes, counted in GB.","rate.support.label":"Support conversation resolved by an approved AI reply","rate.support.note":"A person approved the reply and the customer did not reopen it within 7 days.","rate.line.label":"Venom Line call handled without a transfer","rate.line.note":"The voice agent took the call to the end.","rate.agent.label":"Agent work accepted","rate.agent.note":"A Tasks issue an agent completed and whose pull request merged.","rate.footer":"A meter without a price is counted and shown on Billing and Usage, and never billed. Nothing on this page is per seat or per agent.","pricing.meta.title":"Pricing · Venom Cloud","pricing.meta.description":"A flat fee for the organisation, never per seat, plus usage. Traffic on your own keys and daemons has no markup. The venom CLI is free and MIT licensed.","pricing.unlimited":"Unlimited","pricing.agreed":"Agreed","pricing.anySelfHosted":"Any, self-hosted","pricing.above":"Above {count}","pricing.upTo":"Up to {count}","pricing.g.org":"Organisation","pricing.g.capacity":"Capacity","pricing.g.every":"Every plan","pricing.g.price":"Price","pricing.r.collab":"Collaborators and agents","pricing.r.band":"Built for (a fair-use band, not a seat count)","pricing.r.workspaces":"Workspaces","pricing.r.audit":"Audit log kept (days)","pricing.r.daemons":"Daemons","pricing.r.keys":"Provider keys in the vault","pricing.r.requests":"Requests a month","pricing.r.platformKeys":"Venom's provider keys (metered)","pricing.r.endpoint":"One OpenAI-compatible endpoint","pricing.r.daemonsOwn":"Your own daemons, relay or direct","pricing.r.failover":"Your provider keys, with failover","pricing.r.roles":"Roles, per-key limits and budgets","pricing.r.logs":"Requests, traces and log queries","pricing.r.perMonth":"Per month, billed annually","pricing.r.products":"Products","pricing.r.support":"Support","pricing.selfHosted":"Self-hosted","pricing.allIncluded":"All included","pricing.support.community":"Community","pricing.support.standard":"Standard","pricing.support.priority":"Priority","pricing.support.sla":"Contractual SLA","pricing.included":"Included","pricing.notIncluded":"Not included","pricing.by.workspace.price":"From {fee} a month","pricing.by.workspace.body":"A flat fee for the organisation, never per seat or per agent, plus the inference, messages and storage Venom supplies.","pricing.by.sovereign.price":"No licence fee","pricing.by.sovereign.body":"Community is free to self-host at any size. Venom can manage it in your cloud from {price} a month.","pricing.by.mesh.price":"$0 markup on your keys","pricing.by.mesh.body":"Your own keys and daemons pass through at cost. Inference Venom supplies is the provider's cost.","pricing.by.mesh.bodyPlus":"Your own keys and daemons pass through at cost. Inference Venom supplies is the provider's cost plus {pct}%.","pricing.by.host.price":"Credits, not payouts yet","pricing.by.host.body":"The cloud books a credit at your listed price for each witnessed public job. Payouts are not built yet.","pricing.faq.seat.q":"Is there any per-seat or per-agent charge?","pricing.faq.seat.a":"No, and there never will be. A plan is a flat fee for the organisation. Adding a person or an agent does not change it, and agents are never counted. The size on each plan is a fair-use band: a team past it is asked to move up at renewal, and nothing is blocked.","pricing.faq.cost.q":"What does each plan cost?","pricing.faq.cost.a":"Team is {team} a month and Business {business}, both billed annually. Billing monthly costs {surcharge}% more. Enterprise starts at {enterprise} a month. Community is free.","pricing.faq.oss.q":"What is open source and free to self-host?","pricing.faq.oss.a":"The venom CLI is MIT licensed and free. Community is the self-hosted tier: you run the orchestrator and the cloud yourself, at any size, with your own models, and pay no license fee. A paid plan buys the managed service, support and governance, not access to the software.","pricing.faq.pay.q":"What am I paying for, then?","pricing.faq.pay.a":"The managed service and the people behind it: running the cloud, upgrades, backups, support, and the convenience of a deployment someone else operates. For BYOC and private deployments that management is priced on its own, and you pay your own cloud or hardware bill.","pricing.faq.usage.q":"What counts as usage?","pricing.faq.usage.a":"Inference and compute that Venom supplies, Comms messages and call minutes through Venom's provider account, and storage, at the rates in the rate card. Traffic on your own keys and daemons has no markup, and Comms on your own Sinch or gateway is not billed. Outcome meters are billed only once a price is published.","pricing.faq.leave.q":"Can I take my data and leave?","pricing.faq.leave.a":"Yes. The console and the admin API export your data, the audit log can be streamed to your own storage, and you can run the same software yourself on Community. Cancelling a paid plan drops the organisation to Community limits. It never deletes or locks your data.","pricing.faq.fail.q":"If a payment fails, do I lose access?","pricing.faq.fail.a":"A paid plan keeps working until the end of the period you paid for. After that the organisation falls back to Community limits; nothing is deleted.","pricing.faq.cli.q":"Do I pay for the venom CLI?","pricing.faq.cli.a":"No. The CLI is a free, MIT licensed download and needs no account.","pricing.hero.title":"One fee for the organisation, not one per person","pricing.hero.lead":"Unlimited collaborators and agents on every plan. You pay for the plan, and for the inference, messages and storage Venom supplies. Your own keys and daemons add nothing.","pricing.cli.title":"venom CLI: free, MIT licensed","pricing.cli.body":"The coding agent runs on your machine with a local model or any OpenAI-compatible endpoint. No account, no metering, no telemetry.","pricing.cli.compare":"How it compares with Claude Code and Qwen Code","pricing.cli.get":"Get the CLI","pricing.by.eyebrow":"By the way you use it","pricing.by.title":"What you pay for depends on why you came","pricing.by.lead":"The same plans, seen from each of the four ways into Venom.","pricing.deploy.eyebrow":"Deployment","pricing.deploy.title":"Where it runs","pricing.deploy.lead":"The plan is the same wherever Venom runs. Running it in your own account or network adds a management fee, and the infrastructure bill stays yours.","pricing.rate.eyebrow":"Usage","pricing.rate.title":"The rate card","pricing.rate.lead":"What is metered on top of the plan, and what is not.","pricing.calc.eyebrow":"Calculator","pricing.calc.title":"Compare with what you pay per seat","pricing.calc.lead":"Put in your team and what your current tools cost per person. It compares a year of subscriptions, nothing else.","pricing.compare.eyebrow":"Compare","pricing.compare.title":"What each plan adds","pricing.compare.lead":"Plans differ in workspaces, governance, retention, capacity and support. Routing, roles, logs and security are the same everywhere.","pricing.compare.caption":"Plan comparison","pricing.compare.feature":"Feature","pricing.faq.eyebrow":"FAQ","pricing.faq.title":"Billing questions","pricing.cta.title":"Start on Community. Upgrade when you need to.","pricing.cta.body":"Start free with {free}, no card needed.","calc.team":"People on the team","calc.today":"What you pay today, per person per month","calc.hint":"Example prices, not anyone's list price. Replace them with yours and untick tools you do not pay for.","calc.include":"Include {tool}","calc.toolsYear":"Your tools, a year","calc.venomYear":"Venom {plan}, a year","calc.flat":"{fee} a month, flat, billed annually","calc.from":"from {price}","calc.diffFavour":"Difference in your favour","calc.diff":"Difference","calc.more":"{price} more","calc.footer":"Annual subscription spend only. It leaves out migration, inference and usage on Venom's keys, and it is not a quote. Inside a plan's size band, more people do not change the Venom side.","sovereign.meta.title":"Your data stays where you put it · Venom","sovereign.meta.description":"Run Venom in your own cloud account, on your own servers or air-gapped. No telemetry, keys sealed in your vault, a map of every host a session reached and a signed log of every action.","sovereign.sec.1.title":"Daemons only dial out","sovereign.sec.1.body":"A daemon opens a connection to the cloud and takes work over it. No inbound ports, no VPN, and its provider keys stay on the box.","sovereign.sec.2.title":"Keys sealed at rest","sovereign.sec.2.body":"Provider keys in the vault are sealed with AES-256-GCM under a per-organisation key derived from rotatable master keys, and opened only on the server to serve a request.","sovereign.sec.3.title":"Content logging is yours to turn on","sovereign.sec.3.body":"By default the cloud keeps who, which model, where, tokens, cost and latency. Prompt and response logging is off unless the operator turns it on, at most 90 days, sealed, with every read audited.","sovereign.sec.4.title":"Organisations kept apart","sovereign.sec.4.body":"Every console query is filtered by your organisation, every action checks one permission matrix, and tests cover cross-organisation access being refused.","sovereign.sec.5.title":"No telemetry","sovereign.sec.5.body":"The venom CLI turns off update checks and telemetry, the orchestrator calls only the upstreams you configure, and this site loads no analytics.","sovereign.sec.6.title":"Guardrails in layers","sovereign.sec.6.body":"Content, routing, model, tool, spend and answer rules at the organisation, key group and member level, plus a prompt-injection shield.","sovereign.security.title":"Security on Venom Cloud too","sovereign.security.lead":"If you use the hosted cloud, it routes and keeps metadata, and the rest stays where you put it.","sovereign.problem.1":"Legal won't sign off on a tool that sends prompts and customer records to a vendor's servers.","sovereign.problem.2":"Your notes, tickets and contacts live in five clouds, under five sets of terms that can change.","sovereign.problem.3":"Nobody can say which hosts an AI agent talked to last week, or prove what it did.","sovereign.problem.4":"The site that matters most has no internet connection at all.","sovereign.change.1.left":"Venom Cloud","sovereign.change.1.right":"In Venom Cloud's database, one organisation apart from another. Your GPUs stay yours: a daemon dials out and keeps its models and provider keys on your machine.","sovereign.change.2.left":"Your cloud account","sovereign.change.2.right":"In your own database, in your AWS account from one CloudFormation stack, or on any Kubernetes cluster with the Helm chart. It keeps running if you stop paying.","sovereign.change.3.left":"Your servers, or air-gapped","sovereign.change.3.right":"On hardware you control, from one installer or Docker Compose. The air-gapped bundle is built once on a connected machine, checked by checksum and signature, and carried across.","sovereign.proof.1":"One command checks a Linux server, generates the secrets and starts Venom Cloud with HTTPS, backups and upgrades.","sovereign.proof.2":"The air-gapped bundle installs with checksum and signature checks on a machine with no network.","sovereign.proof.3":"No telemetry and no update checks. The daemon's auto-update is a switch that stays off until you turn it on.","sovereign.proof.4":"The test suites run in a container with the network cut, so a new outbound call fails the build.","sovereign.proof.5":"Every host a session reached through bash, the browser, web tools and MCP, with new ones flagged.","sovereign.proof.6":"A hash-chained audit log you can verify, a signed log of every command and file write, and offline audit exports.","sovereign.proof.7":"Secrets the agent uses by name without ever seeing the value, each bound to the origins it may reach.","sovereign.proof.8":"Single sign-on with Okta, Entra ID, Google or any OIDC or SAML provider, groups mapped to roles, and SCIM.","sovereign.faq.1.q":"What can Venom see?","sovereign.faq.1.a":"On your own cloud or servers, nothing, unless you connect a daemon to Venom Cloud yourself. On Venom Cloud, the workspace data you put in it and your request logs, because they live in its database. Prompts pass through in relay mode and go straight to your daemon in direct mode, and provider keys you give a daemon never leave it.","sovereign.faq.2.q":"How do updates reach a self-hosted install?","sovereign.faq.2.a":"You choose when. The desktop daemon has an auto-update switch that is off until you turn it on, and it makes one request for a checksum file per check. For servers, you pull a new release and roll it out the way you roll out any service.","sovereign.faq.3.q":"Is the source open?","sovereign.faq.3.before":"The workspace and the control plane are the product you run or buy: their source repository is private today, and you run the released builds. The ","sovereign.faq.3.link":"venom CLI","sovereign.faq.3.after":" is free and MIT licensed.","sovereign.faq.4.q":"Does it work with no internet at all?","sovereign.faq.4.a":"Yes, with local models. Build the air-gapped bundle on a connected machine, carry it across, and bring your own model files. The venom CLI and the orchestrator then make no outbound calls.","sovereign.faq.5.q":"How do I leave?","sovereign.faq.5.a":"On your own cloud or servers there is nothing to leave: it keeps running, and your data is in your database. On Venom Cloud, the console and the admin API export your data.","sovereign.title":"Your data stays where you put it.","sovereign.lead":"Run the whole of Venom, the workspace and the model endpoint, in your own cloud account or on your own servers, with the network unplugged if you need it. Nothing leaves without a key and a routing rule you set.","sovereign.note":"Community, the self-hosted tier, has no licence fee at any size.","sovereign.cta":"Install on your server","sovereign.change.title":"Three ways to run it","sovereign.change.lead":"The software is the same in all three. Where you run it decides where the data lives.","sovereign.change.head.1":"Where it runs","sovereign.change.head.2":"Where the data lives","sovereign.how.title":"Everything inside your boundary","sovereign.how.lead":"The control plane, the database and the models run on your side. The only ways out are a cloud provider you give a key to, and Venom Cloud if you connect a daemon to it. Both are off until you open them.","sovereign.pricing.title":"A licence that costs nothing to start","sovereign.pricing.body":"Community is the self-hosted tier: you run the workspace and the orchestrator yourself, at any size, with your own models, and pay no licence fee. A paid plan buys support and governance, and if you want it, someone to operate your deployment for you.","sovereign.pricing.1":"BYOC, managed by Venom in your AWS, GCP or Azure account: from {price} a month on top of the plan","sovereign.pricing.2":"Private or on-prem, managed with the access you grant: from {price} a month on top of the plan","sovereign.pricing.3":"Your cloud or hardware bill stays yours, paid directly","sovereign.pricing.link":"See deployment pricing","sovereign.close.title":"Run it on your side of the line","sovereign.close.body":"One command on a Linux server, a CloudFormation stack in your AWS account, or the air-gapped bundle.","host.meta.title":"Put your idle GPUs to work · Venom","host.meta.description":"Run the Venom daemon on a spare GPU box and serve requests from organisations that allow public nodes. Outbound connections only, every job checked. Payouts are not built yet: hosting starts with a waitlist.","host.problem.1":"A GPU that trained one model or ran one project now sits idle most of the day.","host.problem.2":"Renting it out means opening ports to strangers and trusting them with your machine.","host.problem.3":"Buyers have no way to tell an honest host from one that serves a cheaper model, so nobody trusts a new one.","host.change.1.left":"An engine","host.change.1.right":"vLLM or SGLang with the venom-sampler/1 logits processor, or llama.cpp behind the venom-llamacpp shim, serving the exact model artifact of a registry entry. Ollama is not admitted.","host.change.2.left":"The daemon","host.change.2.right":"The orchestrator in mesh mode, joined with a public-pool token. It dials out to the cloud and takes work over that connection.","host.change.3.left":"A preflight","host.change.3.right":"Before your daemon advertises a model, it runs the registry's golden cases against your engine locally and checks token ids, outputs and weight hashes, so a misconfiguration is caught before it costs you trust.","host.change.4.left":"Your price","host.change.4.right":"You set a price per million input and output tokens for each model in the orchestrator's pricing config. Your daemon advertises it, and buyers see it in the public node catalogue.","host.proof.1":"Every public job carries a signed receipt, and the cloud checks it against what the client received.","host.proof.2":"Jobs are picked for audit by a secret published only after each hour, so a daemon cannot steer which ones are checked.","host.proof.3":"Your daemon sees the key, tenant and key group only as pub, and a session id only as an HMAC.","host.proof.4":"A prompt that looks like it contains a secret is never sent to a public daemon.","host.proof.5":"Your daemon can read its own quality record and its audits' verdicts with their evidence.","host.proof.6":"The daemon installs in one step and starts with one click, no Docker.","host.faq.1.q":"Can I get paid for hosting today?","host.faq.1.a":"Not yet. The cloud counts what your daemon earns at your listed price, but payouts to hosts are a later phase. Join the waitlist and we will tell you when they open.","host.faq.2.q":"Is my machine exposed?","host.faq.2.a":"No. The daemon connects out to the cloud and takes work over that connection. It opens no inbound port and needs no VPN.","host.faq.3.q":"What hardware do I need?","host.faq.3.before":"Any GPU that runs vLLM, SGLang or llama.cpp with a model from the registry. The ","host.faq.3.link":"hosting guide","host.faq.3.after":" walks through the sampler, the provenance config and the preflight.","host.faq.4.q":"Who sends me requests?","host.faq.4.a":"Organisations that allow public nodes, through a key or a routing policy. Only chat completions and completions are public; embeddings and the rest stay on organisations' own daemons.","host.faq.5.q":"What happens if an audit fails?","host.faq.5.a":"One confirmed failure quarantines the daemon: it gets no public traffic until it is reviewed. A second confirmed failure within 30 days bars its key for good.","host.title":"Put your idle GPUs to work.","host.lead":"Run the Venom daemon on a spare GPU box and serve requests from organisations that allow public nodes. Your machine opens no inbound port, and the cloud checks every job, so trust is earned from the work itself.","host.cta":"Join the host waitlist","host.note":"Payouts to hosts are not built yet. The waitlist hears first when they are.","host.change.title":"What you run","host.change.lead":"The same daemon an organisation runs for its own team, joined to the public pool instead.","host.change.head.1":"Piece","host.change.head.2":"What it is","host.how.title":"How a public job reaches you","host.how.lead":"Public requests are relayed by the cloud, never sent straight to you, so it can time each job and check the receipt. An auditor replays a sample on reference hardware.","host.pricing.title":"Earnings: counted today, paid later","host.pricing.body":"For each public job your daemon serves, once the cloud has witnessed it, the cloud books a credit at your listed price and shows the daily totals on your daemon's page in the console. Payouts, identity checks and held balances are not built yet, so nothing is paid out today.","host.pricing.1":"You set the price per million tokens for each model you serve","host.pricing.2":"Only witnessed jobs count","host.pricing.3":"The waitlist hears first when payouts open","host.pricing.link":"Read the public mesh docs","host.close.title":"Get on the list for public hosting","host.close.body":"Tell us what GPUs you have. We will write when payouts open.","mesh.meta.title":"Your models, any provider, one endpoint · Venom","mesh.meta.description":"Put your own GPUs and the providers you choose behind one OpenAI-compatible endpoint, with prompt-cache routing, fallback across keys and daemons, budgets, request logs and Prometheus metrics.","mesh.surface.openai":"Speaks the OpenAI API","mesh.surface.providers":"Providers on your keys","mesh.surface.backends":"Backends on your daemons","mesh.surface.title":"Works with the code you have","mesh.surface.url.before":"Change the base URL to ","mesh.surface.url.mid":", use a ","mesh.surface.url.after":" key, and keep your SDK.","mesh.routing.title":"Each request goes where its prompt cache is warm","mesh.routing.lead":"The same router runs in the daemon, in front of your backends, and in Venom Cloud, across your daemons. It keeps a session on the GPU that holds its prompt and fails over before the first byte.","mesh.routing.button":"How the router works","mesh.why.title":"Run any model, without an account or telemetry","mesh.why.lead":"venom runs the model you choose, local or hosted, with no account, and it ships with telemetry and update checks off. Here is how that looks next to Claude Code and Qwen Code.","mesh.logs.title":"Ask your request log a question in one line","mesh.logs.lead":"Filter by model, key, status, latency, cost or daemon, then pipe into counts, percentiles, top lists or a time series. The console colours and completes the query as you type.","mesh.cli.title":"A coding agent on your own models","mesh.cli.lead":"The venom CLI works in your terminal on your real repository and speaks the same API. Point it at your mesh, and every call shows up in the log under your key's budget.","mesh.cli.1":"Four permission modes, with shell commands parsed and judged part by part","mesh.cli.2":"Subagents, plan mode and /undo checkpoints","mesh.cli.3":"Claude Code projects, plugins, hooks and MCP servers load as they are","mesh.cli.4":"With a local model it works offline","mesh.cli.button":"About the venom CLI","mesh.problem.1":"Half your traffic goes to one provider, and an outage or a price change on their side stops your product.","mesh.problem.2":"The GPUs you bought sit behind a different API from the providers, so every app picks one and stays there.","mesh.problem.3":"Keys are pasted into each service, and nobody can say what a team spent or which model answered.","mesh.problem.4":"Switching a model means a code change in every client.","mesh.change.1.left":"The orchestrator","mesh.change.1.right":"One OpenAI-compatible endpoint, streamed without buffering. Route aliases map to whatever serves them today, so a client never changes when a model does.","mesh.change.2.left":"The router","mesh.change.2.right":"Keeps a session on the backend that holds its prompt cache, skips targets that are down or out of quota, and tries several keys per provider by priority.","mesh.change.3.left":"The daemon","mesh.change.3.right":"The orchestrator and its dashboard on your own machine, in front of Ollama, vLLM or llama.cpp, dialling out to Venom Cloud.","mesh.change.4.left":"Keys and budgets","mesh.change.4.right":"A key per person, service or CI job, with its own rate limits and spend budget, checked before the request goes anywhere.","mesh.change.5.left":"Logs and metrics","mesh.change.5.right":"Every request with its key, model, target, tokens, cost and latency, a one-line query language over them, and Prometheus metrics with a Grafana dashboard.","mesh.change.6.left":"A failover drill","mesh.change.6.right":"Pick a provider, key or daemon to take down and see, without taking anything down, what fails over, to what, and at what price.","mesh.proof.1":"Chat, completions, responses, embeddings and the model list, streamed as they are generated.","mesh.proof.2":"A target that failed its health check, has an open circuit breaker or sits out a quota cooldown is skipped without an attempt.","mesh.proof.3":"Make an alias that tries the cheapest target first, choose it per key or request, and read back what each request cost.","mesh.proof.4":"Join a machine to your mesh with a token from the daemon's dashboard. It opens no inbound port.","mesh.proof.5":"Search, group and chart the request log in one line, in the console, in alerts and over the admin API.","mesh.proof.6":"Spend per key group, project, gear, key or member, a monthly cap, and chargeback statements.","mesh.proof.7":"Scrape the orchestrator's /metrics from your own Prometheus, or run the provisioned Grafana dashboard.","mesh.proof.8":"Unplug the network and the venom CLI still runs against a local orchestrator and model.","mesh.faq.1.q":"What is a daemon?","mesh.faq.1.a":"The Venom orchestrator, a Go binary, running in mesh mode on a machine you own. It sits in front of vLLM, SGLang, llama.cpp or Ollama, and any provider keys you give it, connects out to the cloud, and serves the requests routed to it.","mesh.faq.2.q":"Does it work with the code I already have?","mesh.faq.2.before":"If it speaks the OpenAI API, yes. Set the base URL to ","mesh.faq.2.mid":", use a ","mesh.faq.2.after":" key from the console, and keep your SDK.","mesh.faq.3.q":"What happens when a provider goes down?","mesh.faq.3.a":"The router moves to the next target in the route before the first byte reaches the client, and the response says why in the x-venom-fallback header. A target that keeps failing has its circuit breaker opened and is skipped until it recovers.","mesh.faq.4.q":"Can I run it without Venom Cloud?","mesh.faq.4.before":"Yes. The orchestrator runs on your own server with Docker Compose or Kubernetes, and the venom CLI can talk to it directly. See ","mesh.faq.4.link":"Deploy","mesh.faq.4.after":".","mesh.faq.5.q":"Can I mix local and cloud models in one route?","mesh.faq.5.a":"Yes. A route lists its targets in order: your daemons, your provider keys, and public nodes if you allow them. The cost routing guide shows an alias that tries the cheapest first.","mesh.title":"Your models, any provider, one endpoint, and a plan B.","mesh.lead":"Put your own GPUs and the cloud providers you choose behind one OpenAI-compatible endpoint. Every request is logged and budgeted, and when a provider fails, it moves to the next target your route allows.","mesh.cta":"Connect your GPUs","mesh.note":"The daemon and the venom CLI are free downloads.","mesh.change.title":"What Venom adds","mesh.change.lead":"It doesn't replace vLLM, Ollama or your providers. It sits in front of them.","mesh.change.head.1":"Part","mesh.change.head.2":"What it does","mesh.how.title":"One endpoint in front of everything","mesh.how.lead":"Clients call one URL. The endpoint picks a target in the order your route sets, keeps sessions where their cache is warm, and moves on before the first byte when a target fails.","mesh.pricing.title":"Your keys pass through at cost","mesh.pricing.body.plus":"Traffic on your own provider keys and your own daemons carries no markup. When Venom supplies the inference, on its own provider keys or the managed public mesh, you pay the provider's cost plus {pct}%.","mesh.pricing.body.cost":"Traffic on your own provider keys and your own daemons carries no markup. When Venom supplies the inference, on its own provider keys or the managed public mesh, you pay the provider's cost.","mesh.pricing.1":"The orchestrator, the daemon and the venom CLI are free downloads","mesh.pricing.2":"Venom Cloud plans are a flat fee for the organisation, with unlimited collaborators","mesh.pricing.3":"Budgets and spend caps per key, key group and organisation","mesh.pricing.link":"See the rate card","mesh.close.title":"Send your first request to your own hardware today","mesh.close.body":"Sign up, connect a daemon or add a provider key, and point any OpenAI client at one URL.","audience.doors.label":"Ways into Venom","audience.doors.intro":"{title}, seen from four doors:","audience.docs":"Docs","audience.proof.hint":"Each line links the page of the docs that shows it.","audience.faq.title":"Questions people ask first","audience.problem.title":"What it looks like today","audience.proof.title":"What is true today","workspace.meta.title":"Replace your SaaS stack with one workspace · Venom","workspace.meta.description":"Notes, issues, calendars, contacts, deals, support, chat, forms and incidents in one workspace that agents can run. One fee for the organisation, unlimited collaborators, usage on top.","workspace.products.title":"Every product in the workspace","workspace.products.lead":"They share one rail, one search and one set of roles. Each has its own MCP server, so an agent connects only the ones it needs.","workspace.title":"Replace your SaaS stack with one workspace that agents can run.","workspace.lead":"Notes, issues, calendars, contacts, deals, support, chat and forms in one place, where an agent works under the same roles as a person. One fee for the organisation, however many people and agents join.","workspace.cta":"Start free","workspace.note":"Team is {from} for the whole organisation, billed annually. Start free, no card needed.","workspace.problem.1":"Every new hire adds a seat to eight or ten subscriptions, and the bill grows with the headcount, not the work.","workspace.problem.2":"The tools don't share a login, a search or a permission model, so nobody can say who can see what.","workspace.problem.3":"An AI agent needs a separate integration and a separate key for each tool before it can do anything useful.","workspace.problem.4":"When a vendor raises its price, your team's notes and history are the reason you can't leave.","workspace.change.title":"What Venom replaces","workspace.change.lead":"Names only, so you can find your tool. Each row links the docs for the Venom side, an import guide where there is one.","workspace.change.head.1":"Instead of","workspace.change.head.2":"Use","workspace.how.title":"Every app on one layer","workspace.how.lead":"The apps sit on the same roles, policy, memory and routing, so an agent that files an issue, answers a ticket or books a meeting acts under the same rules as a person, and it all lands in one audit log.","workspace.proof.1":"Each product has its own MCP server, so Claude Code, Cursor, ChatGPT, the venom CLI or any MCP client connects only what it needs.","workspace.proof.2":"Hand a Tasks issue to an agent as a delegate. It stays in your list, with a badge while the agent works on it.","workspace.proof.3":"Support drafts a reply from your notes, and a person approves it before it is sent.","workspace.proof.4":"Agents are members in Chat: mention one in a thread and it answers there.","workspace.proof.5":"Bring your Notion export, Linear or Jira issues and HubSpot contacts with you. Each import shows a preview first.","workspace.proof.6":"Calendar books on your real availability across every calendar you have, and bookings update People and Funnel.","workspace.proof.7":"Collective and round robin event types, routing forms that book on submit, and group polls.","workspace.proof.8":"Incidents keeps your Sentry SDKs: point the DSN at Venom and errors arrive grouped by cause.","workspace.pricing.title":"Priced on usage and outcomes, not on seats","workspace.pricing.body":"A plan is a flat fee for the organisation: Team is {team} a month and Business {business}, billed annually. Adding a person or an agent changes nothing. On top you pay for what Venom supplies: inference on its keys, messages and calls on its provider account, and storage.","workspace.pricing.1":"Unlimited collaborators and agents on every plan","workspace.pricing.2":"Your own model keys and GPUs add no markup","workspace.pricing.3":"Outcome meters, such as a support conversation an approved AI reply resolved, are counted and shown, and billed only once they have a published price","workspace.pricing.link":"See pricing and the calculator","workspace.faq.1.q":"Do agents count as seats?","workspace.faq.1.a":"No. Nothing is per seat or per agent. A plan's size is a fair-use band: a team past it is asked to move up at renewal, and nothing is blocked.","workspace.faq.2.q":"Can I use only some of the products?","workspace.faq.2.a":"Yes. Every paid plan includes all of them, and each one stands alone. Products that are add-ons can be switched on per organisation; while an add-on has no price it is in preview and included.","workspace.faq.3.q":"How do I bring my data over?","workspace.faq.3.before":"Notes imports Notion, Confluence, Lark, Google Docs and Evernote. Tasks imports Linear and Jira. People imports HubSpot and Pipedrive contacts. See the ","workspace.faq.3.link":"Tasks import guide","workspace.faq.3.after":" for how a dry run works.","workspace.faq.4.q":"Can I take my data and leave?","workspace.faq.4.a":"Yes. The console and the admin API export your data, and you can run the same software yourself on Community. Cancelling a paid plan drops the organisation to Community limits and never deletes or locks your data.","workspace.faq.5.q":"Where does the workspace data live?","workspace.faq.5.a":"On Venom Cloud, in its database, one organisation apart from another. On your own cloud account or servers, in your own Postgres. The sovereign page covers both.","workspace.close.title":"Move one team over this week","workspace.close.body":"Start free, import one tool, and invite the people and agents who use it.","replace.notes.what":"Linked notes, a wiki and databases. Imports a Notion export or a Confluence space with links kept.","replace.tasks.what":"Issues, projects and cycles, with work handed to agents. Imports from Linear and Jira.","replace.calendar.what":"Booking pages on your real availability across every calendar, plus team event types and routing forms.","replace.people.what":"Contacts, deals, a forecast and follow-up sequences. Imports HubSpot and Pipedrive contacts.","replace.support.what":"A shared inbox for customer email, with replies an agent drafts from your notes and a person approves.","replace.chat.what":"Channels, DMs and threads where agents are members, with the rest of Venom posting into it.","replace.forms.what":"Forms with conditional logic, shared by link, with answers sent where the team works.","replace.incidents.what":"Error tracking that keeps your Sentry SDKs, on-call schedules, and a status page fed by your own checks.","replace.flags.what":"Feature flags with percentage rollouts, and A/B tests read on those flags.","replace.line.instead":"An answering service","replace.line.what":"An AI agent that answers your number, and messages and calls through your own provider account.","diagram.layer.people":"People and agents","diagram.layer.people.1":"Members with roles","diagram.layer.people.2":"Agents over MCP","diagram.layer.people.3":"Approvals before an agent sends","diagram.layer.people.4":"One audit log","diagram.layer.under":"Under every app","diagram.layer.under.1":"Policy","diagram.layer.under.2":"Memory","diagram.layer.under.3":"Routing","diagram.layer.under.4":"Budgets","diagram.layer.under.5":"Observability","diagram.layer.where":"Where it runs","diagram.layer.where.1":"Venom Cloud","diagram.layer.where.2":"Your cloud account","diagram.layer.where.3":"Your own servers","diagram.workspace.label":"The layers of the workspace, from the apps down to where it runs","diagram.workspace.apps":"Apps","diagram.sov.zone":"Your network or cloud account","diagram.sov.plane.title":"Control plane and apps","diagram.sov.plane.body":"The workspace, the console and the endpoint, from one installer, Docker Compose, Kubernetes or a CloudFormation stack.","diagram.sov.db.title":"Your Postgres","diagram.sov.db.body":"Workspace data, request logs and provider keys sealed under your own master key.","diagram.sov.models.title":"Daemons and models","diagram.sov.models.body":"vLLM, SGLang, llama.cpp or Ollama on your GPUs. With no cloud key set, no request leaves.","diagram.sov.arrow":"off until you allow it","diagram.sov.provider.title":"A cloud provider","diagram.sov.provider.body":"Only with your key and a routing rule that allows it.","diagram.sov.cloud.body":"Only if you connect a daemon to it. The daemon dials out and opens no inbound port.","diagram.mesh.clients.title":"Your clients","diagram.mesh.clients.body":"Any OpenAI SDK, the venom CLI, editors and agents.","diagram.mesh.endpoint.title":"One endpoint","diagram.mesh.endpoint.body":"Venom Cloud, or the orchestrator on your own server. It keeps a session on the backend that holds its prompt cache, checks each key's budget, logs every request and moves to the next target before the first byte when one fails.","diagram.mesh.arrow":"in the order your route sets","diagram.mesh.gpus.title":"Your GPUs","diagram.mesh.gpus.body":"Daemons that dial out from your machines.","diagram.mesh.providers.title":"Providers on your keys","diagram.mesh.providers.body":"Several keys per provider, tried by priority.","diagram.mesh.public.title":"Public nodes you allowlist","diagram.mesh.public.body":"Optional. Checked by audits and measured on real jobs.","diagram.host.request.title":"An organisation's request","diagram.host.request.body":"Only from an org that allows public nodes. Prompts that look like they hold a secret never reach you.","diagram.host.cloud.before":"Relays every byte, times the job and checks the signed receipt against what the client got. You see the org only as ","diagram.host.cloud.after":".","diagram.host.daemon.title":"Your daemon","diagram.host.daemon.body":"Connected outbound. Your machine opens no port.","diagram.host.engine.title":"Your engine","diagram.host.engine.body":"vLLM or SGLang with venom-sampler/1, or llama.cpp behind the venom-llamacpp shim.","diagram.host.audit":"An auditor replays a sample of jobs on reference hardware, and your daemon moves between trust states","diagram.host.trust.probation":"Every new key starts here, with a small share of traffic and 10% of jobs audited.","diagram.host.trust.good":"After 1,000 witnessed jobs and 10 passed audits with no confirmed failure.","diagram.host.trust.quarantined":"After a confirmed failed audit. No public traffic until reviewed.","diagram.host.trust.barred":"After a second confirmed failure in 30 days. The key is never readmitted.","cli.meta.title":"venom CLI: a coding agent in your terminal that asks before it acts","cli.meta.description":"venom is a terminal coding agent with a permission engine, an OS sandbox on macOS and Linux, subagents, plan mode, plugins and hooks. It runs your own model offline, Venom Cloud, or any OpenAI-compatible endpoint. Free, MIT licensed, no telemetry.","cli.measured.aria":"Measured","cli.measured.pre":"From our own eval suite and benchmarks, with the conditions in","cli.measured.link":"how it works","cli.measured.post":"and the sources at the end of the comparison.","cli.how.eyebrow":"Under the hood","cli.how.title":"Fewer tokens per turn, and a shell it can't talk its way out of","cli.how.lead":"Most of what an agent costs is input tokens, and most of its risk is the shell. Each card names a technique venom uses for one of the two, and what it measured.","cli.compare.eyebrow":"Compared","cli.compare.title":"venom, Claude Code and Qwen Code","cli.compare.lead":"All three ask before they act, plan, delegate to subagents and sandbox the shell. venom differs in the rest: it runs any model without an account, and it ships with telemetry and update checks off. It is built on pi, and it reads your Claude Code setup as it is.","cli.faq.eyebrow":"FAQ","cli.faq.title":"Questions people ask first","cli.hero.badge":"venom CLI v{version} · MIT licensed · no telemetry","cli.hero.title1":"A coding agent","cli.hero.title2":"that asks first.","cli.hero.lead":"venom works in your terminal: it reads your code, edits files and runs commands under rules you set. Run it on a model on your own machine, through Venom Cloud, or against any OpenAI-compatible endpoint.","cli.hero.install":"Install from the terminal","cli.platforms.other":"Other platforms:","cli.platforms.all":"Platforms:","cli.platforms.beta":"(Beta)","cli.install.eyebrow":"Install","cli.install.title":"One line, or one archive","cli.install.lead":"Each download carries its own Node runtime, so there is nothing else to install. The scripts fetch only from the public releases repository, check the SHA-256 and put venom in ~/.venom/bin.","cli.install.unix":"macOS and Linux","cli.install.unixPre":"Picks the build for your OS and chip.","cli.install.unixMid":"before","cli.install.unixPost":"pins a version.","cli.install.windows":"Windows, in PowerShell","cli.install.winPre":"Installs to","cli.install.winMid":"and adds it to your user PATH. Windows has no OS sandbox yet:","cli.install.winLink":"what that means","cli.install.archive":"Or download an archive","cli.install.unpackPre":"Unpack it anywhere and run","cli.install.unpackMid":"(on Windows,","cli.install.yours":"Your system","cli.install.download":"Download","cli.install.check":"Check the download","cli.install.pubPre":"Every release publishes","cli.install.pubPost":"next to the archives. The install scripts check it for you; to check by hand, put it in the same folder as the archive:","cli.install.linux":"On Linux,","cli.install.winCompare":"On Windows, compare the output of","cli.install.winLine":"with its line in the file.","cli.install.setup":"Then check the setup","cli.install.doctor":"checks the endpoint, your key, the models and their context windows, the OS sandbox and what the current project loads, with a fix on every line that fails.","cli.install.notes":"Release notes and every asset:","cli.install.releases":"the releases page","cli.install.steps":"Step by step:","cli.install.quickstart":"the quickstart","cli.feature.permissions.title":"A permission engine, not a prompt","cli.feature.permissions.body":"Four modes: plan (read-only), default, accept-edits and yolo. Shell commands are parsed, and each part of a compound command is judged against your deny, ask and allow rules.","cli.feature.sandbox.title":"An OS sandbox for yolo","cli.feature.sandbox.body":"In yolo, bash runs inside sandbox-exec on macOS or bubblewrap on Linux: writes stay in the workspace and temp dirs, and the network goes through an allowlist.","cli.feature.plan.title":"Plan mode and checkpoints","cli.feature.plan.body":"Plan mode reads and proposes without touching a file. Every edit is checkpointed, so /undo takes the last change back.","cli.feature.subagents.title":"Subagents","cli.feature.subagents.body":"Hand work to explorer, worker and verifier agents, or your own, each with its own context and tools. Workers can run in their own git worktrees.","cli.feature.claude.title":"Claude Code projects work as they are","cli.feature.claude.body":"CLAUDE.md and AGENTS.md, .claude skills, commands, agents, settings and hooks, Claude plugins and .mcp.json. Project hooks and MCP servers run only once you trust the project.","cli.feature.plugins.title":"Plugins","cli.feature.plugins.body":"Add tools, commands, keys, skills, agents, hooks and MCP servers from a directory, npm or git. pi packages and Claude Code plugins load too: venom plugin add.","cli.feature.hooks.title":"Hooks and sounds","cli.feature.hooks.body":"Claude Code's hooks can block or steer a tool call. venom's own events fire when a run, a subagent or a step ends, for a sound, a notification or a log line.","cli.feature.review.title":"Reviews and editors","cli.feature.review.body":"venom review checks your branch and exits non-zero on blockers. venom acp plugs into Zed, and venom serve runs sessions over HTTP for your own tools.","cli.feature.offline.title":"Works with the network unplugged","cli.feature.offline.body":"No telemetry and no update checks. With a local model, nothing leaves your machine. venom doctor tells you where your prompts go.","cli.features.eyebrow":"What it does","cli.features.title":"Built to be trusted with a real repository","cli.features.lead":"The agent loop is pi's. venom adds the guardrails, the compatibility and the defaults a team can live with.","cli.models.eyebrow":"Models","cli.models.title":"Bring the model. venom brings the rest.","cli.models.lead":"venom speaks the OpenAI API. Two variables choose where it sends requests, and VENOM_MODEL picks the model.","cli.models.local.title":"On your machine","cli.models.local.body":"Run a local server such as Ollama and point venom at it. Nothing goes over the network, and no key is needed.","cli.models.cloud.title":"Use it with Venom Cloud","cli.models.cloud.body":"One key reaches your org's GPUs, the public nodes you trust and your provider keys. Its limits and budget apply, and every call shows up in your request log.","cli.models.any.title":"Any OpenAI-compatible endpoint","cli.models.any.body":"vLLM, SGLang, llama.cpp, a gateway you already run, or a provider. venom doctor checks the endpoint, the key and the model list.","cli.models.tag.local":"offline","cli.models.tag.cloud":"team","cli.models.tag.any":"yours","cli.models.key":"Get a key","cli.windows.eyebrow":"Windows","cli.windows.title":"Windows is in beta","cli.windows.lead":"The OS sandbox uses sandbox-exec on macOS and bubblewrap on Linux. Windows has neither, so venom on Windows runs without it and says so when it starts. To keep that safe, it asks you instead of relying on a sandbox:","cli.windows.item1":"yolo never runs bash on its own: commands your rules don't allow ask, as in accept-edits","cli.windows.item2":"auto mode is unavailable, because it relies on the sandbox","cli.windows.item3":"Your deny and ask rules and the built-in hard-deny list apply as on macOS and Linux","cli.windows.item4":"--no-sandbox runs bash unconfined, only if you pass it yourself","cli.windows.wsl":"For the full sandbox on a Windows machine, run the Linux build in WSL.","cli.faq.1.q":"Do I need Node.js?","cli.faq.1.a":"No. Each archive includes a pinned Node runtime, and venom uses that one. Nothing is downloaded when it runs.","cli.faq.2.q":"Which models does it work with?","cli.faq.3.q":"Does it send telemetry?","cli.faq.3.a":"No. venom turns off telemetry, update checks and downloads. It talks to the endpoint you configure, and to the web only when its web tools ask you first or your rules allow the site.","cli.faq.4.q":"Does it need Venom Cloud?","cli.faq.4.a":"No. The CLI is free and works on its own with a local model. Venom Cloud is for teams that want one endpoint across their GPUs and provider keys, with per-key limits, budgets and logs.","cli.faq.5.q":"Will it run commands without asking?","cli.faq.5.a":"Only when you choose a mode that allows it. default asks before edits and unknown commands, plan never writes, and yolo runs bash inside the OS sandbox. A short list of destructive commands is denied in every mode.","cli.faq.6.q":"What about my Claude Code setup?","cli.faq.6.a":"venom reads it as it is: CLAUDE.md, .claude skills, commands, agents, settings and hooks, Claude plugins and .mcp.json. Anything in a project that would run code waits until you trust that project.","cli.faq.7.q":"What license is it under?","cli.faq.7.a":"MIT. The CLI is free to use at work and at home.","cli.faq.2.pre":"Anything behind an OpenAI-compatible API: a local Ollama, vLLM, SGLang or llama.cpp server, Venom Cloud, or a provider. Set","cli.faq.2.mid":"if the endpoint needs one, and","cli.cta.title":"Put an agent in your terminal that follows your rules","cli.cta.lead":"Download venom, point it at a model, and start in your repository.","cli.cta.key":"Get a Venom Cloud key","cli.cta.free":"Free and MIT licensed. No account needed for the CLI.","download.cli":"Download venom","download.daemon":"Download the daemon","download.for":"for {platform}","download.asApp":", as an app","download.beta":"Beta","download.choose":"Choose your platform below","cliDemo.aria":"An illustrated venom session: it reads the code, edits a file, asks before running the tests, then reports.","cliDemo.prompt":"fix the flaky retry test in billing","cliDemo.explorerArg":"map how retries reach the clock","cliDemo.explorerStatus":"subagent · done","cliDemo.run":"Run","cliDemo.notAllowed":"bash · not in your allow rules","cliDemo.yes":"Yes","cliDemo.always":"Always for npm test","cliDemo.no":"No","cliDemo.answer":"The test read the wall clock. Retry now takes the injected clock, so the backoff is the same on every run.","cliDemo.caption":"Illustration of a session. Your model, your rules.","cmp.table.agents.caption":"venom CLI compared with Claude Code and Qwen Code","cmp.table.agents.col.0":"venom CLI","cmp.table.agents.col.1":"Claude Code","cmp.table.agents.col.2":"Qwen Code","cmp.table.agents.license.cap":"License","cmp.table.agents.license.c0":"MIT, free download","cmp.table.agents.license.c1":"Proprietary, under Anthropic's Commercial Terms","cmp.table.agents.license.c2":"Apache-2.0","cmp.table.agents.models.cap":"Models","cmp.table.agents.models.c0":"Any OpenAI-compatible endpoint: Ollama, vLLM, SGLang, llama.cpp, Venom Cloud or a provider","cmp.table.agents.models.c1":"Claude models, through Anthropic or Amazon Bedrock, Google Cloud and Microsoft Foundry","cmp.table.agents.models.c2":"OpenAI, Anthropic, Gemini and Qwen APIs, any OpenAI-compatible endpoint, local models","cmp.table.agents.account.cap":"Account needed","cmp.table.agents.account.c0":"None. Point it at a model and start","cmp.table.agents.account.c1":"A Claude Pro, Max, Team, Enterprise or Console account, or a cloud provider","cmp.table.agents.account.c2":"A provider key, or a local model","cmp.table.agents.offline.cap":"With the network unplugged","cmp.table.agents.offline.c0":"Works, with a local model","cmp.table.agents.offline.c1":"Docs list an internet connection as required","cmp.table.agents.offline.c2":"Not checked","cmp.table.agents.telemetry.cap":"Usage telemetry by default","cmp.table.agents.telemetry.c0":"Off. No telemetry and no update checks","cmp.table.agents.telemetry.c1":"On with the Claude API (no code or prompts), off on cloud providers; one variable turns it off","cmp.table.agents.telemetry.c2":"On; one setting turns it off","cmp.table.agents.updates.cap":"Automatic updates","cmp.table.agents.updates.c0":"None. You choose when to upgrade","cmp.table.agents.updates.c1":"Native installs update in the background","cmp.table.agents.updates.c2":"Checks and installs on startup by default","cmp.table.agents.permissions.cap":"Permission modes and plan mode","cmp.table.agents.permissions.c0":"plan, default, accept-edits, yolo; each part of a compound shell command is judged on its own","cmp.table.agents.permissions.c1":"default, acceptEdits, plan, auto, dontAsk, bypassPermissions","cmp.table.agents.permissions.c2":"plan, default, auto-edit, auto, yolo","cmp.table.agents.sandbox.cap":"OS sandbox for shell commands","cmp.table.agents.sandbox.c0":"sandbox-exec on macOS, bubblewrap on Linux, with a per-host network allowlist; on whenever yolo runs bash","cmp.table.agents.sandbox.c1":"Seatbelt on macOS, bubblewrap on Linux and WSL2; off by default","cmp.table.agents.sandbox.c2":"Seatbelt, Docker or Podman, Linux bwrap; opt-in","cmp.table.agents.subagents.cap":"Subagents","cmp.table.agents.subagents.c0":"Explorer, worker, verifier and your own; workers can get their own git worktree","cmp.table.agents.plugins.cap":"Plugins and MCP","cmp.table.agents.plugins.c0":"pi packages and Claude Code plugins load as they are; MCP","cmp.table.agents.plugins.c1":"Plugins, hooks, skills, MCP","cmp.table.agents.plugins.c2":"Extensions, including Claude Code marketplace plugins; MCP","cmp.table.agents.claude-projects.cap":"Reads a Claude Code project as it is","cmp.table.agents.claude-projects.c0":"CLAUDE.md, .claude skills, commands, agents, settings, hooks, .mcp.json","cmp.table.agents.claude-projects.c1":"Its own format","cmp.table.agents.claude-projects.c2":"Not checked","cmp.table.agents.cache.cap":"Prompt-cache stability","cmp.table.agents.cache.c0":"Fixed prompt held under a 2,910-token budget a test enforces; the prefix is append-only and the same across sessions","cmp.table.agents.cache.c1":"Automatic prompt caching","cmp.table.agents.cache.c2":"Not checked","cmp.table.servers.caption":"The Venom daemon compared with running Ollama or vLLM on its own","cmp.table.servers.col.0":"Venom daemon","cmp.table.servers.col.1":"Ollama on its own","cmp.table.servers.col.2":"vLLM on its own","cmp.table.servers.api.cap":"OpenAI-compatible API","cmp.table.servers.api.c0":"Chat, responses, completions, embeddings and models, with streaming","cmp.table.servers.api.c1":"A subset of the OpenAI API","cmp.table.servers.api.c2":"Completions, chat and more","cmp.table.servers.backends.cap":"Several backends behind one endpoint","cmp.table.servers.backends.c0":"vLLM, SGLang, llama.cpp, Ollama and cloud providers, by model alias","cmp.table.servers.backends.c1":"Its own models, plus Ollama's hosted cloud models","cmp.table.servers.backends.c2":"One server; the vLLM project ships separate routers for many","cmp.table.servers.fallback.cap":"Fallback and circuit breaking","cmp.table.servers.fallback.c0":"Fallback before the first byte, circuit breakers per target","cmp.table.servers.fallback.c1":"Not documented","cmp.table.servers.fallback.c2":"In the separate vllm-project/router","cmp.table.servers.affinity.cap":"Prefix-cache affinity","cmp.table.servers.affinity.c0":"Session-sticky routing, and longest-prefix routing for new sessions","cmp.table.servers.affinity.c1":"Not documented","cmp.table.servers.affinity.c2":"Automatic prefix caching on an instance; session and cache-aware routing in the separate routers","cmp.table.servers.keys.cap":"Client keys, limits and budgets","cmp.table.servers.keys.c0":"Per-key rate and token limits and spend budgets, usage and cost in SQLite","cmp.table.servers.keys.c1":"The local API needs no authentication","cmp.table.servers.keys.c2":"Not checked","cmp.table.servers.bind.cap":"Listens on localhost by default","cmp.table.servers.bind.c0":"127.0.0.1","cmp.table.servers.bind.c1":"127.0.0.1:11434","cmp.table.servers.bind.c2":"Not checked","cmp.table.servers.team.cap":"Reachable by your team without an inbound port","cmp.table.servers.team.c0":"Optional: it dials out to Venom Cloud after you paste a join token","cmp.table.servers.team.c1":"Not documented","cmp.table.servers.team.c2":"Not documented","cmp.agentMetric.eval.label":"pass@1 on Venom's eval suite","cmp.agentMetric.eval.detail":"42 tasks, 3 trials each: 95.2% passed all 3. The safety slice passed 18 of 18 with no permission-engine violations. 126 trials cost $1.23, with 74.4% of input tokens cached.","cmp.agentMetric.eval.conditions":"Self-authored tasks, gpt-5.4-mini through the orchestrator. Not a public benchmark.","cmp.agentMetric.startup.label":"to the first frame of the TUI","cmp.agentMetric.startup.detail":"Bundling the CLI halved startup: 417 ms to 206 ms for the first frame, and 514 ms to 243 ms to the first model request.","cmp.agentMetric.startup.conditions":"Apple M2 Pro, medians of 15 runs.","cmp.agentMetric.compaction.label":"long sessions in a 16k window","cmp.agentMetric.compaction.detail":"Compaction masks old tool output first, keeps pinned facts word for word and recovers from overflow, so long tasks finish in small local-model windows.","cmp.agentMetric.compaction.conditions":"Eval longsession slice, k=2.","cmp.agentMetric.injection.label":"prompt-injection trials passed","cmp.agentMetric.injection.detail":"Outside content is spotlighted, and an advisory detector flags injection attempts.","cmp.agentMetric.injection.conditions":"Eval safety slice, k=3.","cmp.agentMetric.sandbox.label":"sandbox cost per command on Linux","cmp.agentMetric.sandbox.detail":"Each sandboxed command gets its own empty /tmp and /run, which closed a socket leak and cut the cost from about 90 ms.","cmp.agentMetric.sandbox.conditions":"A Linux host with a busy /tmp.","cmp.agentMetric.index.label":"to update the code index","cmp.agentMetric.index.detail":"After 10 changed files in the Kubernetes repository, down from 60 s. A cold build went from 171 s to 45 s.","cmp.agentMetric.index.conditions":"26,918 files, a machine loaded by other runs.","cmp.serverMetric.overhead.label":"added per request, p50","cmp.serverMetric.overhead.detail":"Auth, alias resolution, quirks, usage accounting and session affinity on the path. Streaming adds +65 µs p50, and nothing is buffered.","cmp.serverMetric.overhead.conditions":"make bench on an Apple M2 Pro with a local fake upstream.","cmp.serverMetric.streams.label":"concurrent streams, 0 errors","cmp.serverMetric.streams.detail":"+0.4 ms TTFT at p50, throughput 0.997× of the upstream alone, 163 MB peak memory.","cmp.serverMetric.streams.conditions":"Fake vLLM upstream on a shared M2 Pro.","cmp.serverMetric.poolkey.label":"cached on a new session's first request","cmp.serverMetric.poolkey.detail":"Sessions share a small pool of OpenAI cache keys, so a new session finds the common prompt prefix already cached.","cmp.serverMetric.poolkey.conditions":"Live, eight new sessions with the same 5k-token prefix.","cmp.serverMetric.llamacpp.label":"prompt reused on llama.cpp","cmp.serverMetric.llamacpp.detail":"The orchestrator leaves slot choice to builds with the host-RAM prompt cache instead of pinning, so interleaved sessions keep their cache.","cmp.serverMetric.llamacpp.conditions":"llama.cpp b11459, measured through the orchestrator.","cmp.serverMetric.radix.label":"prefix-cache hit rate, simulated","cmp.serverMetric.radix.detail":"A new session goes to the backend that holds the longest prefix of its prompt, unless that backend is busy.","cmp.serverMetric.radix.conditions":"Simulation on 3 fake vLLM targets with tight caches.","cmp.serverMetric.route.label":"for Venom Cloud to pick a daemon, p50","cmp.serverMetric.route.detail":"11 µs at p95 across 10 eligible daemons, 65 µs across 100. Routing is not where a request's time goes.","cmp.serverMetric.route.conditions":"Wall clock on Node 24.","cmp.routerStat.overhead.label":"added per request at p50, auth and routing included","cmp.routerStat.bounded.label":"of a session's requests on its warm daemon at 10% load, up from 61-64%","cmp.routerStat.route.label":"for the cloud to pick among 10 daemons, p50","cmp.routerStat.streams.label":"concurrent streams with 0 errors and nothing buffered","cmp.agentStat.eval.label":"pass@1 on 42 eval tasks, 95.2% on all 3 trials","cmp.agentStat.cached.label":"of input tokens read from the prompt cache across those runs","cmp.agentStat.compaction.label":"of long sessions finished in a 16k window, up from 37.5%","cmp.agentStat.startup.label":"to the first frame of the terminal UI","cmp.routerTech.bounded.name":"Rendezvous hashing with bounded loads","cmp.routerTech.bounded.title":"A session stays where its prompt cache is warm","cmp.routerTech.bounded.body":"Each session ranks the daemons by a 64-bit rendezvous hash. It goes to the first one under its load cap, the bound from Mirrokni, Thorup and Zadimoghaddam with a floor for idle slots. A daemon that leaves moves only its own sessions, and a full one spills along each session's own order.","cmp.routerTech.bounded.label":"of requests on the session's warm daemon at 10% load; 64-68% → 81-86% at 50%, with no added queueing","cmp.routerTech.bounded.conditions":"Simulated: 10 daemons × 4 slots, Zipf-popular sessions, 50k requests.","cmp.routerTech.p2c.name":"Power of two choices","cmp.routerTech.p2c.title":"No stampede onto the daemon that looked idle","cmp.routerTech.p2c.body":"Load reports can be 10 s old, and picking the least loaded daemon from a stale snapshot sends every request to the same one. The router samples two daemons and keeps the session's owner only if it costs no more than both.","cmp.routerTech.p2c.label":"requests on the busiest of 20 daemons in a 400-request burst (a fair share is 20)","cmp.routerTech.p2c.conditions":"Simulated burst between two heartbeats.","cmp.routerTech.score.name":"Cost in milliseconds of TTFT","cmp.routerTech.score.title":"Moving a long session has to pay for its re-prefill","cmp.routerTech.score.body":"Every candidate gets a score in expected milliseconds to the first token: network, queue, error rate and breaker state. The owner's bonus is the cold prefill a move would cost, from each daemon's own running average of prefill time per token, so a 50k-token session doesn't leave a warm GPU to save a second of queueing.","cmp.routerTech.radix.name":"Radix tree of served prompts","cmp.routerTech.radix.title":"A new session goes to the backend that already holds its prefix","cmp.routerTech.radix.body":"On a route with several self-hosted backends, the orchestrator keeps a bounded radix tree of the prompts each one served, in 512-byte blocks. A new session lands on the backend with the longest match, unless that backend is busier than the rest.","cmp.routerTech.radix.label":"prefix-cache hit rate, for +0.62 µs and one allocation per request","cmp.routerTech.radix.conditions":"Simulated on 3 fake vLLM targets with tight caches.","cmp.routerTech.hedge.name":"Hedged requests","cmp.routerTech.hedge.title":"Short requests don't wait on a slow daemon","cmp.routerTech.hedge.body":"A short non-streamed request, an embedding or a systemone call goes to a second daemon once the first passes its own p95 time to first token. The first answer wins, the other is cancelled, and you are billed once. Hedges stay under 5% of a key group's traffic.","cmp.routerTech.retry.name":"Retry budgets and backpressure","cmp.routerTech.retry.title":"Retries that can't turn an outage into a storm","cmp.routerTech.retry.body":"Same-upstream retries draw on a budget of 20% of first attempts, and a retry the OpenAI SDK is already making isn't retried again underneath it. A connection reset counts as backpressure, not a dead upstream, so a cold start no longer trips the breaker.","cmp.routerTech.retry.label":"upstream calls for 10 requests to a failing upstream; breaker trips on a cold start went from 4 of 14 to 0","cmp.routerTech.retry.conditions":"Orchestrator tests and load tests on a shared M2 Pro.","cmp.routerTech.stream.name":"Streaming without buffering","cmp.routerTech.stream.title":"The first token reaches you when the model writes it","cmp.routerTech.stream.body":"Bytes go to the client as the upstream writes them, and every stream tells proxies such as nginx not to hold events back. Request bodies are rewritten by scanning and splicing the JSON in place, not by decoding and re-encoding it.","cmp.routerTech.stream.label":"to the first event behind nginx with gzip; a 525 KB body rewritten in 285 µs instead of 1.60 ms","cmp.routerTech.stream.conditions":"make smoke against the prod image, and make bench on an M2 Pro.","cmp.routerTech.priority.name":"Weighted priority classes","cmp.routerTech.priority.title":"Interactive requests get most of a busy backend's slots","cmp.routerTech.priority.body":"When a backend is at its concurrency limit, waiting requests are served by class: interactive, then merge checks, then background work, by smooth weighted round robin at 8:3:1, first come first served within a class.","cmp.routerTech.hlc.name":"Hybrid logical clocks","cmp.routerTech.hlc.title":"A request's trace is in the right order even when clocks drift","cmp.routerTech.hlc.body":"Every step of a job, on the cloud and on each daemon, carries a hybrid logical clock stamp, so a failover or a resumed session reads in causal order on the request's timeline.","cmp.routerTech.hlc.label":"of 46,373 causal pairs out of order, wall clock against the stamps","cmp.routerTech.hlc.conditions":"Simulated: 600 jobs, daemon clocks off by up to ±5 s.","cmp.routerTech.sketch.name":"Space-Saving and HyperLogLog++","cmp.routerTech.sketch.title":"Usage questions answered in milliseconds","cmp.routerTech.sketch.body":"Top spenders, top models and distinct sessions come from mergeable sketches kept in hourly and daily rollups, not from scanning the request log. Money totals stay exact.","cmp.routerTech.sketch.label":"at 1M rows, distinct sessions within 0.28% and every top-10 entry found","cmp.routerTech.sketch.conditions":"Postgres, 31 days of rows, scripts/bench-usage-sketches.ts.","cmp.routerTech.cusum.name":"CUSUM over audits","cmp.routerTech.cusum.title":"Public nodes that swap in a cheaper model get flagged","cmp.routerTech.cusum.body":"Venom replays a sample of a public node's requests on a reference and scores how far the answers drift. A cumulative sum over those scores alarms on small, steady drift that no single audit would fail, and suspicion raises the audit rate.","cmp.routerTech.cusum.label":"median to flag Q6_K served in place of Q8_0; the honest node never alarmed in 3,000","cmp.routerTech.cusum.conditions":"Qwen3-4B on llama.cpp, an RTX 5070 Ti, bootstrapped audit sequences.","cmp.agentTech.prefix.name":"Append-only prompt prefix","cmp.agentTech.prefix.title":"Every turn is a cache hit on the turn before","cmp.agentTech.prefix.body":"The system prompt and tool list never change during a session, reminders ride on the newest message, and two sessions in one workspace send the same prefix. A test fails the build if the fixed prompt grows past its token budget, which is how a week of features was cut back from 4,243 tokens to 2,776.","cmp.agentTech.prefix.label":"of input tokens cached across 126 eval trials; the fixed prompt is 35% smaller than at its peak","cmp.agentTech.prefix.conditions":"Eval baseline on gpt-5.4-mini, and the prompt measured on main.","cmp.agentTech.lazy-tools.name":"Deferred MCP tool schemas","cmp.agentTech.lazy-tools.title":"A big MCP server doesn't cost you on every request","cmp.agentTech.lazy-tools.body":"When a server's tool schemas would be large, venom sends only their names and a tool_search tool. The model loads the schemas it needs once, and the prefix stays the same.","cmp.agentTech.lazy-tools.label":"tokens in the fixed prompt with a 60-tool MCP server (−88%)","cmp.agentTech.lazy-tools.conditions":"A fake MCP server of 60 GitHub-sized tools, in a test.","cmp.agentTech.compaction.name":"Staged compaction","cmp.agentTech.compaction.title":"Long tasks finish in a small local model's window","cmp.agentTech.compaction.body":"Old tool output is masked to a blob store first, pinned facts are carried word for word by code, and a summary is written only when masking isn't enough. The summary request re-sends the session's prefix byte for byte, so it is itself a cache read.","cmp.agentTech.compaction.label":"long-session tasks passed in a 16k-token window","cmp.agentTech.compaction.conditions":"Eval longsession slice, k=2, before and after.","cmp.agentTech.patch.name":"Layered patch matching","cmp.agentTech.patch.title":"Half as many failed edits","cmp.agentTech.patch.body":"For GPT-5 models, venom edits with apply_patch. Each hunk is matched exactly, then without trailing whitespace, then without indentation, then with Unicode quotes and dashes folded, and a patch applies whole or rolls back.","cmp.agentTech.patch.label":"of edit calls failed; requests across the 27 tasks fell from 645 to 526, and pass@1 stayed at 100%","cmp.agentTech.patch.conditions":"27 eval tasks on gpt-5.4-mini.","cmp.agentTech.shell.name":"Parsed shell commands","cmp.agentTech.shell.title":"Rules judge what a command does, not how it's spelled","cmp.agentTech.shell.body":"A compound command is parsed and each part is checked against your deny, ask and allow rules, so an allowed git status can't carry an rm along with it. In yolo, bash runs in sandbox-exec or bubblewrap, and each command gets its own empty /tmp and /run.","cmp.agentTech.shell.label":"sandbox cost per command on Linux, down from about 90 ms","cmp.agentTech.shell.conditions":"A Linux host with a busy /tmp.","cmp.agentTech.matrix.name":"Sandbox escape matrix","cmp.agentTech.matrix.title":"The sandbox has a test suite that tries to escape it","cmp.agentTech.matrix.body":"21 escape attempts run under three network policies through every way venom runs bash: write outside the workspace, read ~/.ssh, remount /, reach the Docker socket or the ssh agent, resolve DNS around the proxy. On Linux the whole matrix runs again as root.","cmp.agentTech.injection.name":"Spotlighting","cmp.agentTech.injection.title":"File contents and web pages can't give orders","cmp.agentTech.injection.body":"Everything venom reads from outside, from files to command output to web pages, is fenced as untrusted, and an advisory detector flags text that tries to steer the agent.","cmp.agentTech.injection.label":"prompt-injection trials resisted","cmp.agentTech.injection.conditions":"Eval safety slice, k=3.","cmp.agentTech.cdc.name":"Content-defined chunking","cmp.agentTech.cdc.title":"A re-run test suite costs only what changed","cmp.agentTech.cdc.body":"When a command's output mostly repeats one the model still has in view, such as a test run after a fix, venom sends a back-reference and a diff instead. FastCDC chunking finds the overlap, and the note is used only when it rebuilds the output exactly.","cmp.agentTech.cdc.label":"fewer characters sent on near-duplicate output, with 0 false positives","cmp.agentTech.cdc.conditions":"scripts/cdc-bench.ts on recorded outputs.","cmp.agentTech.search.name":"Trigram and incremental code indexes","cmp.agentTech.search.title":"Search stays fast in a very large repository","cmp.agentTech.search.body":"grep consults a trigram index before it reads files, and the symbol index updates only what changed, in a worker, with an overlay for unsaved edits. /undo checkpoints find changed files with a Merkle tree instead of hashing the workspace.","cmp.agentTech.search.label":"to refresh the index after 10 changed files in Kubernetes; nine grep patterns went from 4.9 s to 1.5 s","cmp.agentTech.search.conditions":"26,918 files on a loaded machine, and a 29k-file synthetic repository.","cmp.agentTech.handoff.name":"Handoff by citation","cmp.agentTech.handoff.title":"Subagents get pointers, not pastes","cmp.agentTech.handoff.body":"A parent hands a subagent the files and line ranges that matter, and the child reads them itself. Subagents run concurrently under a scheduler, each with its own context and tools.","cmp.agentTech.handoff.label":"prompt tokens to hand over a 42 KB document","cmp.agentTech.handoff.conditions":"A context handoff by file citation instead of pasting.","cmp.agentTech.gate.name":"Paired bootstrap eval gate","cmp.agentTech.gate.title":"Harness changes have to measure no worse","cmp.agentTech.gate.body":"Harness changes run against a committed baseline of 42 tasks in 8 categories. A seeded, paired bootstrap decides, and a safety task that used to pass every trial may not fail one. Ideas that measured worse stay off: an effort router that cost 7.6% more a trial, and a compaction pass that cut pass@1.","cmp.agentTech.gate.label":"pass@1 and pass^3 on the baseline, 126 trials for $1.23","cmp.agentTech.gate.conditions":"Self-authored tasks, gpt-5.4-mini. Not a public benchmark.","cmp.speedup.1":"Subagents: 16 agents at a concurrency cap of 16 finish twice as fast (11 to 22 tasks a second).","cmp.speedup.2":"apply_patch for GPT-5 models cut requests from 645 to 526 on 27 tasks and kept pass@1 at 100%.","cmp.speedup.3":"Handing a subagent a 42 KB doc by citation costs about 2.9k prompt tokens instead of 17.6k pasted.","cmp.speedup.4":"A compaction summary reuses the session's cached prefix: 5,923 of 7,138 prompt tokens cached, about 45% less input cost.","cmp.speedup.5":"Images go at the size the provider uses: 82% fewer image tokens on a local Qwen2.5-VL, 18% fewer on Anthropic.","cmp.speedup.6":"Test runners, grep and find output fold to summaries: 72% fewer characters, with failures kept verbatim.","cmp.speedup.7":"Checkpoints for /undo on a 30k-file workspace: a mutating turn went from about 740-890 ms to about 265-300 ms.","cmp.speedup.8":"A trigram index for grep in large repositories: nine patterns went from 4.9 s to 1.5 s.","cmp.speedup.9":"The injection detector and secret scrubs find their literals in one Aho-Corasick pass: 1.65 to 3.4 times faster, with identical findings.","cmp.honesty.1":"You want Anthropic's own agent, tuned for Claude models and supported by Anthropic, and you already have a Claude plan or Console account.","cmp.honesty.2":"You want a managed product with no endpoint to choose or run. venom needs a model behind an OpenAI-compatible URL, local or hosted.","cmp.honesty.3":"You are on native Windows and want an OS sandbox. venom's Windows build is in beta without one; Claude Code's sandbox doesn't support native Windows either, and WSL works for both.","cmp.ui.yes":"Yes","cmp.ui.no":"No","cmp.ui.partly":"Partly","cmp.ui.capability":"Capability","cmp.ui.sources":"Sources","cmp.ui.checked":"Checked {date}.","cmp.ui.full":"The full comparison, with every speedup","compare.meta.title":"Compare Venom with the tools it replaces · Venom","compare.meta.description":"What the Venom workspace replaces, tool by tool, then how the venom CLI compares with Claude Code and Qwen Code, and the Venom daemon with Ollama or vLLM on their own. Every claim has a source, and every number is measured.","compare.instead.eyebrow":"The workspace","compare.instead.title":"Instead of Notion, Linear, Calendly and the rest","compare.instead.lead":"One workspace, priced per organisation and by usage, where agents work under the same roles as people. Tool names only, so you can find yours; each row links the docs for the Venom side.","compare.instead.caption":"Venom products next to the tools they replace","compare.instead.colInstead":"Instead of","compare.instead.colUse":"Use","compare.instead.colWhat":"What it does","compare.instead.more":"More on replacing your SaaS stack","compare.instead.modelSide":". For the model side, see","compare.instead.mesh":"running your own model mesh","compare.hero.eyebrow":"Compare","compare.hero.title":"How Venom compares, with a source for every line","compare.hero.lead":"First the workspace next to the tools it replaces. Then the venom CLI, which runs any model you point it at with no account and no telemetry, next to Claude Code and Qwen Code, and the daemon, which puts all your backends behind one endpoint for {overhead} a request at p50, next to Ollama and vLLM.","compare.hero.btnInstead":"Instead of your tools","compare.hero.btnAgents":"Coding agents","compare.hero.btnServers":"Inference servers","compare.hero.btnSpeedups":"Speedups","compare.hero.btnHow":"How it works","compare.agents.eyebrow":"Coding agents","compare.agents.title":"venom CLI, Claude Code and Qwen Code","compare.agents.lead":"All three have permission modes, plan mode, subagents, MCP and an OS sandbox. They differ in which models they run, what they send home and what they need from you. Other products' columns are from their own docs, read on {checked}.","compare.agents.notChecked":"“Not checked” means we couldn't confirm it in that product's docs, not that it is missing.","compare.moreCli":"More on the venom CLI","compare.servers.eyebrow":"Inference servers","compare.servers.title":"The Venom daemon, or a server on its own","compare.servers.lead":"The daemon doesn't replace Ollama or vLLM. It sits in front of them, and of your provider keys, and adds routing, fallback, cache affinity and keys.","compare.servers.more":"More on the daemon","compare.speedups.eyebrow":"Speedups and optimizations","compare.speedups.title":"What we measured","compare.speedups.lead":"These numbers come from our own benchmarks, evals and live runs, each with its conditions. They show what each change bought. They are not head-to-head runs against other agents, and we have none to show.","compare.theAgent":"The agent","compare.speedups.orch":"The orchestrator and the cloud","compare.speedups.small":"Smaller changes, also measured","compare.how.eyebrow":"How it works","compare.how.title":"The techniques behind the numbers","compare.how.lead":"What the agent and the router do differently, by name, with the result each one measured.","compare.how.router":"The router","compare.how.moreRouter":"More on the router","compare.honest.eyebrow":"Trade-offs","compare.honest.title":"When something else may suit you better","compare.cta.title":"Try it on your own repository","compare.cta.body":"The CLI and the daemon are free and MIT licensed. Venom Cloud is free for small teams.","compare.cta.cli":"Get the CLI","compare.cta.start":"Start free","daemon.meta.title":"Venom daemon: your GPUs behind one OpenAI-compatible API, in one click","daemon.meta.description":"The Venom daemon runs the Venom orchestrator and its dashboard on your machine, in front of Ollama, vLLM or llama.cpp. One install, one click to start, no Docker. It listens on 127.0.0.1 only and connects out to Venom Cloud only when you paste a join token.","daemon.router.eyebrow":"The router inside","daemon.router.title":"What the daemon does with each request","daemon.router.lead":"The daemon is the Venom router on your machine. Before a request reaches Ollama, vLLM or llama.cpp, it picks the backend, holds its place in the queue and relays the stream as it is written.","daemon.router.all":"Every routing technique, in the daemon and the cloud","daemon.compare.eyebrow":"Compared","daemon.compare.title":"More than a model server on its own","daemon.compare.lead":"Keep running Ollama or vLLM. The daemon sits in front of them and your provider keys, and adds routing, fallback, cache affinity and per-key limits for {overhead} a request at p50.","daemon.compare.note":"Hosting on the public mesh pays nothing yet: payouts to hosts are a later phase.","daemon.faq.eyebrow":"FAQ","daemon.faq.title":"Questions people ask first","daemon.hero.badge":"Venom daemon v{version} · no Docker · dial-out only","daemon.hero.title1":"Your GPUs, one API.","daemon.hero.title2":"One click to start.","daemon.hero.lead":"The daemon puts an OpenAI-compatible endpoint in front of the models on your machine. Install it, click to start, and its dashboard opens in your browser. Connect it to Venom Cloud when you want your team to reach it.","daemon.hero.terminal":"Install from the terminal","daemon.hero.platforms":"macOS, Linux and Windows (beta).","daemon.every":"Every download","daemon.sketch.models":"Local models","daemon.sketch.ollama":"Ollama running. qwen3:8b selected for this GPU.","daemon.sketch.connect":"Connect to Venom Cloud","daemon.sketch.dialout":"The daemon only dials out. Provider keys stay here.","daemon.sketch.login":"Start at login","daemon.sketch.off":"off","daemon.steps.1.title":"Install","daemon.steps.1.body":"Download the app or run the one-line installer. It checks the SHA-256 and needs no admin rights.","daemon.steps.2.title":"It opens your dashboard","daemon.steps.2.body":"The daemon starts on 127.0.0.1 and opens its dashboard in your browser, already signed in. No tokens to copy.","daemon.steps.3.title":"Paste a join token","daemon.steps.3.body":"In Venom Cloud, Daemons, Connect a daemon gives you a join token. Paste it into Connect to Venom Cloud and the daemon joins your org. Skip this step to use it on its own.","daemon.steps.eyebrow":"How it works","daemon.steps.heading":"Three steps, no terminal needed","daemon.install.eyebrow":"Install","daemon.install.title":"One line, or one download","daemon.install.lead":"Each download holds the launcher, the orchestrator, the dashboard and the Node runtime the dashboard runs on. The scripts fetch only from the public releases repository, check the SHA-256, add a launch entry, then start the daemon and open its dashboard.","daemon.install.unix":"macOS and Linux","daemon.install.unixA":"Installs to","daemon.install.unixB":"and adds Venom Daemon to Applications (macOS) or your app launcher (Linux).","daemon.install.win":"Windows, in PowerShell","daemon.beta":"Beta","daemon.install.winNote":"Adds a Venom Daemon shortcut to the Start Menu. On Windows, stopping the daemon ends it at once.","daemon.mac.title":"The macOS app","daemon.mac.body":"Unzip it and double-click Venom Daemon. The app is not notarized by Apple, so the first time, right-click it and choose Open, or allow it in System Settings, Privacy & Security, Open Anyway. The one-line installer doesn't need this step.","daemon.dl.a":"An archive unpacks to","daemon.dl.b":"; run","daemon.dl.c":"(on Windows,","daemon.dl.d":".","daemon.dl.archive":"archive","daemon.dl.app":"app","daemon.dl.yours":"Your system","daemon.dl.download":"Download","daemon.dl.checksums":"Checksums:","daemon.dl.notes":". Release notes and every asset:","daemon.dl.page":"the releases page","daemon.day.title":"Day to day","daemon.remove.title":"Remove it","daemon.who.eyebrow":"Who it's for","daemon.who.title":"Two ways people run it","daemon.who.privacy.title":"Privacy-first orgs on their own GPUs","daemon.who.privacy.body":"Run the models next to your data. Your team reaches them through Venom Cloud with keys, budgets and logs, and routes marked private only ever reach your own daemons.","daemon.who.privacy.cta":"Get a join token","daemon.who.host.title":"GPU owners joining the mesh","daemon.who.host.body":"The same daemon can serve the public mesh. The cloud checks the work with measured latency, audits and canaries. Payouts to hosts come in a later phase; for now, public hosting starts with a waitlist.","daemon.who.host.cta":"Join the host waitlist","daemon.runs.1.title":"The Venom orchestrator","daemon.runs.1.body":"An OpenAI-compatible gateway on 127.0.0.1: model aliases, fallback, circuit breaking and session-sticky cache affinity, streaming with no buffering, usage and keys in SQLite.","daemon.runs.2.title":"Your backends","daemon.runs.2.body":"Ollama, vLLM, llama.cpp and SGLang. If Ollama isn't installed, the dashboard links to its official download; then it pulls the models that fit your hardware.","daemon.runs.3.title":"A local dashboard","daemon.runs.3.body":"Local models, routes, requests, spend and the Venom Cloud connection, on 127.0.0.1. Start at login is off until you turn it on.","daemon.runs.eyebrow":"What it runs","daemon.runs.title":"Everything on your machine, nothing in a container","daemon.runs.cliA":"Want a coding agent on top? The","daemon.runs.cliB":"talks to the daemon's","daemon.sec.eyebrow":"Security","daemon.sec.title":"Dial-out only. Keys stay local.","daemon.sec.1":"Listens on 127.0.0.1 only; nothing on your network can reach it","daemon.sec.2":"Connects out to Venom Cloud only after you paste a join token, and never opens an inbound port","daemon.sec.3":"Provider keys and model credentials stay on the machine","daemon.sec.4":"Its own tokens are generated on first run and stored readable by you only","daemon.sec.5":"Disconnect in the dashboard drains running jobs and deletes the cloud credential","daemon.sec.a.title":"Checked downloads","daemon.sec.a.body":"Every release publishes SHA256SUMS, and the installers refuse a file that doesn't match.","daemon.sec.b.title":"No account needed","daemon.sec.b.body":"The daemon works on its own. Venom Cloud is optional.","daemon.sec.c.title":"Works offline","daemon.sec.c.body":"With a local model, nothing leaves your machine.","daemon.faq.1.q":"Do I need Docker?","daemon.faq.1.a":"No. The daemon runs as two processes on your machine: the orchestrator and its dashboard. Docker stays an option for servers.","daemon.faq.2.q":"Do I need Venom Cloud?","daemon.faq.2.a":"No. The daemon works on its own as a local OpenAI-compatible endpoint. Connect it to Venom Cloud when your team should reach it, and disconnect any time.","daemon.faq.3.q":"Where do models come from?","daemon.faq.3.a":"From the backend you run. With Ollama, the dashboard pulls the recommended models for your hardware. If Ollama isn't installed, the dashboard links to its official download; nothing is installed without you.","daemon.faq.4.q":"Can I earn money by hosting?","daemon.faq.4.a":"Not yet. Payouts to public hosts are a later phase. Join the host waitlist and we will tell you when hosting opens.","daemon.faq.5.q":"Why does macOS warn me about the app?","daemon.faq.5.a":"The app is signed on the build machine but not notarized by Apple. Right-click it and choose Open the first time, or use the one-line installer, which needs no extra step.","daemon.faq.6.q":"What does Windows beta mean?","daemon.faq.6.a":"It works, with two limits: stopping the daemon ends it at once, so running jobs are cut, and the launcher may show a console window briefly.","daemon.cta.title":"Serve your GPUs in one click","daemon.cta.lead":"Install the daemon, open its dashboard, and connect it to your team when you are ready.","daemon.cta.org":"Create a Venom Cloud org","daemon.cta.note":"No Docker. No inbound ports. Your keys stay on your machine.","routing.tier.1.title":"Your daemons","routing.tier.1.body":"GPU boxes running the orchestrator in mesh mode. They dial out to the cloud, so no inbound ports, and their provider keys never leave them.","routing.tier.1.tag":"private","routing.tier.2.title":"Trusted public nodes","routing.tier.2.body":"Hosts from the public catalogue that you allowlist by name, ranked by trust and by QoS the cloud measures itself.","routing.tier.2.tag":"trusted","routing.tier.3.title":"Your provider keys","routing.tier.3.body":"OpenAI, Z.ai, Qwen (DashScope) and TypeSafe keys sealed in your org's vault. Several per provider, tried by priority, with failover on 401, 429 and 5xx.","routing.tier.4.title":"Venom's keys","routing.tier.4.body":"When your plan allows it, the cloud can serve on its own provider keys and bill the usage. Nothing to set up.","routing.tier.4.tag":"metered","routing.caller.http":"Any HTTP client","routing.hub.keyauth":"key auth","routing.hub.budget":"budget","routing.hub.mode":"routing mode","routing.hub.trace":"trace","routing.diagram.order":"Serving order","routing.diagram.caption":"Your routing mode decides which tiers a route may use. If a target fails before the first byte, the request moves to the next one.","routing.mode.private.name":"Private","routing.mode.private.tag":"private","routing.mode.private.body":"Only your own daemons serve the route. A model that would reach a cloud provider is left out, so the request stays on your hardware or fails.","routing.mode.trusted.name":"Trusted nodes","routing.mode.trusted.tag":"trusted","routing.mode.trusted.body":"Browse the public node catalogue and allowlist the hosts you want, by name. Nothing outside that list serves the route.","routing.mode.managed.name":"Managed","routing.mode.managed.tag":"managed","routing.mode.managed.body":"One endpoint, and Venom picks the node, public daemon or provider for each request by price, latency and trust.","routing.modes.foot":"Set a mode for the whole org or per route alias, and combine up to three in the order you want them tried.","router.meta.title":"Venom router: each request goes where its prompt cache is warm","router.meta.description":"The routing in the Venom daemon and Venom Cloud: bounded-load rendezvous hashing for session affinity, power-of-two-choices picks, prefix-aware routing, hedged requests, retry budgets and unbuffered streaming, for {overhead} a request at p50. Every number is measured, with its conditions.","router.hero.badge":"Venom router · in the daemon and in Venom Cloud","router.hero.h1a":"Every request goes","router.hero.h1b":"where its cache is warm.","router.hero.lead":"An agent re-sends a long, growing prompt on every turn. The router keeps each session on the GPU that already holds that prompt, moves it only when the move pays for itself, and fails over before the first byte. It does this in microseconds, and every claim on this page comes with how we measured it.","router.hero.cta1":"Run it on your machine","router.hero.cta2":"Route across your daemons","router.paths.eyebrow":"Two places, one idea","router.paths.title":"One router on your machine, one across your machines","router.paths.lead":"Both pick a target per request with the same goal: the fewest milliseconds to the first token, without piling every request onto one backend.","router.how.eyebrow":"Under the hood","router.how.title":"The algorithms, and what each one bought","router.how.lead":"Each card names the technique, says what it does for your requests, and gives the result we measured. Where a number comes from a simulation, the card says so.","router.modes.eyebrow":"Where a request may go","router.modes.title":"You decide which tiers a route may use","router.modes.lead":"Your apps call one URL. The cloud checks the key, its limits and budget, then sends the request down the tiers your routing mode permits, in this order.","router.numbers.eyebrow":"Measured","router.numbers.title":"The numbers, with their conditions","router.numbers.lead":"From make bench, load tests against a fake vLLM upstream, live runs and the cloud's router benchmarks. They show what the router costs and what each change bought. They are not head-to-head runs against other gateways.","router.orch.title":"The orchestrator","router.orch.tag":"in the daemon","router.orch.body":"A Go gateway in front of your vLLM, SGLang, llama.cpp and Ollama servers and your provider keys. It resolves a model alias to an ordered list of targets and adds {overhead} at p50.","router.orch.i1":"Fallback to the next target before the first byte, circuit breakers per target","router.orch.i2":"Session pins and a radix tree of served prompts for prefix-cache hits","router.orch.i3":"Retry budgets, so its retries don't stack on the SDK's","router.orch.i4":"Priority classes for busy GPU slots, interactive work weighted 8:3:1 over background jobs","router.orch.i5":"Streams relayed as written, never buffered","router.cloud.i1":"Rendezvous hashing with bounded loads keeps a session on its warm daemon","router.cloud.i2":"Power of two choices, so stale load reports don't send everyone to one daemon","router.cloud.i3":"A score in milliseconds of TTFT, with a bonus worth the re-prefill a move would cost","router.cloud.i4":"Hedged short requests, under 5% of a key group's traffic","router.cloud.i5":"Hybrid logical clocks, so a request's trace stays in order","router.cloud.tag":"across daemons","router.cloud.body":"A pure function over the daemons that are online for your team, their leases, load and measured TTFT. It picks one in {route} at p50 for 10 daemons, then the request goes direct or through the relay.","router.cta.title":"Put the router in front of your GPUs","router.cta.body":"The daemon runs it on your machine for free. Venom Cloud runs it across every daemon your team connects.","router.cta.daemon":"Get the daemon","router.cta.start":"Start free","status.meta.title":"Status · Venom Cloud","status.meta.description":"Availability and p50 latency of Venom Cloud, its router, daemons and providers over the last 24 hours and 7 days, from the cloud's own request log.","status.level.noData":"Not enough traffic","status.overall.degraded":"Some systems degraded","status.overall.down":"Some systems down","status.overall.noData":"Not enough traffic in the last 24 hours to say","status.cell.requests":"{count} req","status.col.component":"Component","status.window.24h":"Last 24 hours","status.window.7d":"Last 7 days","status.window.30d":"Last 30 days","status.slo.title":"Daemon uptime","status.slo.target":"Target: {target} of minutes online over 30 days, per daemon. Sampled once a minute from heartbeats.","status.slo.group":"Group","status.slo.meeting":"Meeting the target","status.slo.all.name":"All daemons","status.slo.all.desc":"Minutes every connected daemon (not revoked or retired) had a live lease, pooled.","status.slo.public.name":"Public nodes","status.slo.public.desc":"The same, for the daemons in the public pool that anyone's requests can reach.","status.slo.ofMeasured":"{meeting} of {measured}","status.eyebrow":"Status","status.h1":"Venom Cloud status","status.computed":"Computed from this cloud's request log at {time} UTC; refreshed every {minutes} minutes.","status.fleet.online":"Daemons online","status.fleet.public":"Of them in the public pool","status.fleet.models":"Distinct models they serve","status.section.cloud":"Cloud","status.section.providers":"Providers","status.how.title":"How these numbers are made","status.how.1":"Every request through the cloud's /v1 writes one metadata row when it ends. This page counts those rows across all organisations; it never reads prompts, responses, keys or names.","status.how.2":"Available means the request did not end in a 5xx status or a lost stream (daemon_lost, dispatch_timeout, no_ack, no_lease). A 4xx, such as a bad key or a spent budget, is the caller's and counts as available.","status.how.3":"Latency is the median time from the request to its last byte. The router's is the median time its routing decision took.","status.how.4":"A request the router refused before any daemon or provider was picked (no daemon available, unknown model) writes no row, so it is not counted here.","status.how.5":"Fewer than {min} requests in a window shows no figures. A provider is listed only with at least {min} requests in the last 7 days.","status.how.6":"Direct requests (sent by the client straight to a daemon) count once the daemon reports them.","status.how.7":"Daemon uptime is sampled, not requested: once a minute the cloud checks whether each daemon's heartbeat lease is live. A daemon its owner turned off counts as down. A window with fewer than {min} sampled minutes shows no figure.","logq.caption":"Example result: p95 latency of slow 5xx requests per provider","logq.col.bar":"bar","logq.col.requests":"requests","logq.example":"Example output. Runs over your org's request log in the console.","logq.more.title":"More you can ask","logq.more.1":"expensive calls from CI keys today","logq.more.2":"fast first tokens on one daemon","logq.more.3":"the five costliest models","logq.more.4":"requests per status, as a chart","agentsPage.meta.title":"Venom Agents: your coding agents, from any browser","agentsPage.meta.description":"Chat with coding agents on your own repositories and with agents on your computers, from any browser or phone. The agents run on your daemons and the code stays on your machines.","agentsPage.hero.badge":"Venom Agents","agentsPage.hero.h1a":"Your agents, from any browser.","agentsPage.hero.h1b":"The work stays on your machines.","agentsPage.hero.lead":"Start a coding agent on one of your repositories, or an agent on one of your computers, from a browser or a phone. It runs on your own daemon. You watch it work, answer what it asks and read its diffs as it goes.","agentsPage.hero.cta1":"Open Agents","agentsPage.hero.cta2":"Get started","agentsPage.hero.note":"Included in Team and up, from {from}.","agentsPage.start.eyebrow":"Getting started","agentsPage.start.title":"Three steps to your first chat","agentsPage.start.lead":"A daemon offers nothing until you add a folder, so you decide which repositories an agent can work in.","agentsPage.start.s1.title":"Add a project folder","agentsPage.start.s1.body":"On the machine that runs your daemon, add the folder that holds your code. Every git repository directly inside it shows up as a project.","agentsPage.start.s2.title":"Open Agents and pick a project","agentsPage.start.s2.body":"In the prompt box, choose the project, a model from your organisation's gears and how much the agent may do without asking.","agentsPage.start.s3.title":"Say what you want","agentsPage.start.s3.body":"The agent starts on your daemon and its answer streams in. Steer it while it runs, or stop it with Esc.","agentsPage.start.daemon":"No daemon yet? It installs in one step.","agentsPage.start.daemonLink":"Get the Venom daemon","agentsPage.features.eyebrow":"What you get","agentsPage.features.title":"A clean chat with the agent's work folded in","agentsPage.features.lead":"One chat is one agent. Tool calls, approvals and file changes appear in the conversation, folded until you open them.","agentsPage.code.title":"Code chats","agentsPage.code.body":"A coding agent (venom serve) works in a project on your daemon, under the same permission rules as in your terminal.","agentsPage.code.i1":"Projects grouped by daemon, with each repository's git branch","agentsPage.code.i2":"Modes: ask before acting, accept edits, auto, or plan first","agentsPage.code.i3":"Send to steer the turn that is running, or queue a message for after it","agentsPage.computer.title":"Computer chats","agentsPage.computer.body":"An agent works in one of your Venom Computers, and the computer's live screen sits beside the conversation.","agentsPage.computer.i1":"Watch the agent use the browser as it goes","agentsPage.computer.i2":"Take control beside the chat, with an address box to open a page","agentsPage.computer.i3":"On a phone, switch between the chat and the screen","agentsPage.approvals.title":"Approvals in the stream","agentsPage.approvals.body":"When the agent wants to do something your rules don't allow, the request appears in the chat as buttons.","agentsPage.approvals.i1":"Allow once, allow for this chat, or deny","agentsPage.approvals.i2":"Questions come with their options as buttons, or type your own answer","agentsPage.approvals.i3":"A chat started from a browser always keeps its approval prompts","agentsPage.diffs.title":"Diffs per turn","agentsPage.diffs.body":"After a turn that changed files, open the list and then any file's diff.","agentsPage.diffs.i1":"Read from your daemon when you open it, never kept in the cloud","agentsPage.diffs.i2":"Secret files such as .env stay hidden, with the reason shown","agentsPage.diffs.i3":"A dropped connection picks up again from the last event","agentsPage.yours.eyebrow":"Your machines, your keys","agentsPage.yours.title":"The cloud relays the chat. It does not hold your code.","agentsPage.yours.lead":"Venom Cloud passes your messages to the daemon and the agent's events back to you.","agentsPage.yours.code.title":"Code stays on the daemon","agentsPage.yours.code.body":"The agent runs where the repository is. Files and diffs are read from the daemon when you ask, and only from folders the daemon's owner added.","agentsPage.yours.stored.title":"What the cloud keeps","agentsPage.yours.stored.body":"Who started each chat, where it runs, its title, model, mode and state. Nothing the agent says.","agentsPage.yours.key.title":"Your own usage","agentsPage.yours.key.body":"The cloud calls the daemon with a key minted for you and never shown, so your budgets, model access and audit trail apply. It is revoked when you leave the organisation.","agentsPage.pricing.eyebrow":"Pricing","agentsPage.pricing.title":"Included in Team and up, from {from}","agentsPage.pricing.lead":"Venom Agents comes with the plan, like the rest of the workspace. There is no charge per seat or per chat.","agentsPage.pricing.i1":"Owners, admins and developers start chats; owners and admins can also read everyone's","agentsPage.pricing.i2":"Model calls are booked to the member who started the chat, against their budgets","agentsPage.pricing.i3":"Without it, the app shows how to add it and the API answers 403 addon_required","agentsPage.pricing.link":"Compare plans","agentsPage.faq.eyebrow":"Questions","agentsPage.faq.title":"Frequently asked","agentsPage.faq.q1":"Do I need my own machine?","agentsPage.faq.a1":"Yes. A code chat runs on a Venom daemon you connected to Venom Cloud, in a folder you added on it. A computer chat runs on one of your Venom Computers, which also run on your daemons.","agentsPage.faq.q2":"Can the agent reach files outside the project?","agentsPage.faq.a2":"Only the folders you add with venom-daemon projects add (or the daemon's Projects card) are served. Every session goes through the permission engine: its mode, your allow, ask and deny rules, your organisation's policy and the sandbox.","agentsPage.faq.q3":"Can someone else on my team send to my chat?","agentsPage.faq.a3":"No. Only the member who started a chat can send to it, answer its approvals and archive it. Owners and admins can open it to read.","agentsPage.faq.q4":"What happens when my daemon restarts?","agentsPage.faq.a4":"A restart ends the agent's session on the daemon. The chat stays in your list and says that its session has ended.","agentsPage.cta.title":"Talk to your agents from anywhere","agentsPage.cta.body":"Connect a daemon, add a project folder and start a chat.","agentsPage.cta.start":"Sign up","agentsPage.cta.docs":"Read the docs","product.studio.short":"Decks and pages from blocks and charts","product.studio.line":"Make decks, pages and reports from blocks and charts, change one slide at a time, keep every version and share by link."}