The shortest path from zero to talking to the agent.
Pick one:
Linux / macOS (one-liner):
curl -fsSL https://raw.githubusercontent.com/zeroclaw-labs/zeroclaw/master/install.sh | bashHomebrew (macOS, Linux):
brew install zeroclawWindows:
Run setup.bat from the latest release, or see Setup β Windows.
From source:
cargo install --locked --path . # inside a clonezeroclaw onboardThe wizard asks ~9 questions. Minimum inputs:
- An LLM provider (Anthropic, OpenAI, Ollama, OpenRouter, etc.) and its API key or endpoint
- At least one channel β the default
clichannel works; add Discord, Telegram, Slack, etc. if you want to chat from those platforms
Everything else has safe defaults. Total time: ~2 minutes.
zeroclaw agentThis drops you into an interactive session using the cli channel. Type, get replies. Pass -m "one-shot message" for a single non-interactive turn.
For always-on deployment, register the service:
zeroclaw service install
zeroclaw service startThen use a chat platform channel to reach the agent from Discord, Telegram, or wherever you configured.
Run with defaults and skip channel setup:
zeroclaw onboard --quick --provider ollama --model qwen3.6:35b-a3bOr go all the way and use YOLO mode β one config preset that disables approvals and safety gates. For dev boxes and home labs only.
- Multi-model setup β fallback chains, routing rules
- Setup β Service management β running as a daemon
- Channels β Overview β wiring up chat platforms
- Security β Autonomy levels β what the agent is allowed to do