Skip to content

quick-start.md

Latest commit

 

History

History
76 lines (48 loc) · 2.07 KB

quick-start.md

File metadata and controls

76 lines (48 loc) · 2.07 KB

Quick Start

The shortest path from zero to talking to the agent.

Install

Pick one:

Linux / macOS (one-liner):

curl -fsSL https://raw.githubusercontent.com/zeroclaw-labs/zeroclaw/master/install.sh | bash

Homebrew (macOS, Linux):

brew install zeroclaw

Windows:

Run setup.bat from the latest release, or see Setup → Windows.

From source:

cargo install --locked --path . # inside a clone

Onboard

zeroclaw onboard

zeroclaw onboard walks through configured sections (model providers, risk profiles, channels, agents, …) and prompts for each. Minimum inputs:

  1. An LLM provider (Anthropic, OpenAI, Ollama, OpenRouter, etc.) and its API key or endpoint
  2. At least one channel — the default cli channel works; add Discord, Telegram, Slack, etc. if you want to chat from those platforms

Everything else has safe defaults. Total time: ~2 minutes.

Talk to it

zeroclaw agent -a <alias>

<alias> matches your [agents.<alias>] config entry — required, no default. This drops you into an interactive session using the cli channel. Pass -m "one-shot message" for a single non-interactive turn.

For always-on deployment, register the service:

zeroclaw service install
zeroclaw service start

Then use a chat platform channel to reach the agent from Discord, Telegram, or wherever you configured.

If onboarding's questions annoy you

Run non-interactively with --quick:

zeroclaw onboard --quick --model-provider ollama --model qwen3.6:35b-a3b

Or go all the way and use YOLO mode — one config preset that disables approvals and safety gates. For dev boxes and home labs only.

Next