Getting started
Pick where to start with venom, whether you use the CLI, run your own models, use Venom Cloud or build an app on the harness.
Venom is the agent-native operating layer you own: a workspace of apps for people and agents, priced on usage and outcomes instead of seats, a coding agent, and the model infrastructure under them, on models and machines you choose. It runs in Venom Cloud, in your cloud or on your servers; Choose how to run Venom compares them, and Products lists every product. The coding agent works with no one else's servers, and every piece works on its own:
| Piece | What it is | Start here |
|---|---|---|
| venom CLI | A coding agent for your terminal, under permission rules you set, with an OS sandbox on macOS and Linux. | venom CLI quickstart |
| Venom Cloud | One OpenAI-compatible endpoint for your own GPUs, trusted public nodes and provider keys, with keys, budgets and logs per key group. | Venom Cloud quickstart |
| Orchestrator | A self-hosted OpenAI-compatible gateway in Go in front of vLLM, SGLang, llama.cpp, Ollama and cloud providers. | Local or cloud models, then the Orchestrator API |
| Desktop daemon | The orchestrator and a dashboard on your own machine, without Docker, for local models. | Local or cloud models |
Harness and venom serve |
The agent harness as a library, and the same sessions over HTTP with Server-Sent Events, for building your own agent apps. | Harness server API |
The path through this section
- Install venom on macOS, Linux or Windows.
- Local or cloud models: point venom at Ollama, your own orchestrator, Venom Cloud or a provider.
- venom CLI quickstart: a first session, permission modes and
venom doctor. - Your first gear: switch between cheap, strong and local models in one keystroke.
- Troubleshooting when something doesn't work.
The shortest version, for a machine with Ollama:
curl -fsSL https://raw.githubusercontent.com/elishahbiz/venom-releases/main/install.sh | sh
export PATH="$HOME/.venom/bin:$PATH"
ollama pull qwen3:8b
export VENOM_BASE_URL=http://localhost:11434/v1
cd ~/work/app && venomBuilding an app on venom
The harness that runs venom's agent loop, permissions, subagents and Claude Code compatibility is a separate library, so you can build your own product on it (a support desk, a collections workflow, an internal tool) without the terminal UI. Two ways in:
- In Node: embed
@venom/harnessand drive sessions from your code. The Harness section covers it. - From any language: run
venom serveand use its HTTP API with Server-Sent Events; there is a Python client. The generated reference is at/docs/api/harness-server.
Either way the models come from the same endpoint choices as the CLI: see Local or cloud models.
For agents
Every page here is also plain markdown: add .md to its URL (for example /docs/getting-started/cli-quickstart.md). /docs/llms.txt lists every page with a one-line description, and /docs/llms-full.txt has all of them in one file.
In this section
- Choose how to run Venom: Compare Venom Cloud, your own cloud and your own servers or an air-gapped site, where the data lives in each, what can leave, and where the install guides are.
- venom CLI quickstart: Install the venom CLI, point it at a model, run a first session and check the setup with venom doctor.
- Venom Cloud quickstart: Sign up, create an API key, install the venom CLI, point it at Venom Cloud and test it.
- Install venom: Install, update and remove the venom CLI on macOS, Linux and Windows, from the release installer, by hand, or from source.
- Local or cloud models: Choose where venom's model runs (Ollama on this machine, the desktop daemon, your own orchestrator, Venom Cloud or any OpenAI-compatible provider) and point the CLI at it.
- Your first gear: Switch between cheap, normal, strong and local models in one keystroke with gears, build a gear for your budget with the gear wizard, and make it the default.
- Troubleshooting: Diagnose a venom setup with venom doctor and fix the common install, endpoint, key, model, sandbox and trust problems.