Onboarding Guide

What happens on first launch

1. CLI asks which provider you want to use.
2. If that provider needs a key, it asks for the API key.
3. CLI tries to fetch live models for that provider, tells you when it is using the live list, and falls back to the built-in catalog if the provider list is unavailable.
4. CLI asks for your default model and agent profile.
5. CLI can optionally configure Telegram, offer to open the Telegram guide if you still need a bot token, and enable TUI gateway autostart.
6. Settings are saved to ~/.waterbrother/config.json (including trace + auto-compact defaults).

npm install -g @tritard/waterbrother
waterbrother --approval on-request

Pick provider and set credentials

1. Choose a provider such as OpenAI, Anthropic, OpenRouter, Ollama, or xAI.
2. Create an API key if that provider requires one.
3. Set provider and key locally:

waterbrother config set provider openai
waterbrother config set apiKey YOUR_PROVIDER_API_KEY

Environment checks

After onboarding, you can still inspect the provider's live model list directly.

waterbrother doctor
waterbrother models status
waterbrother models list
waterbrother models catalog
waterbrother mcp list

Save and reapply provider/model setups

waterbrother runtime-profiles save review-anthropic
waterbrother runtime-profiles list
waterbrother runtime-profiles apply review-anthropic

Prepare remote-control channels

The first-run wizard can offer Telegram setup directly. You can still use service-specific onboarding for Telegram, Discord, and Signal later.

waterbrother onboarding telegram
waterbrother onboarding discord
waterbrother onboarding signal
waterbrother channels status
waterbrother gateway status

Project-local tool servers

Partners can attach repo-local MCP stdio servers through project config.

waterbrother config set-json mcpServers '{"filesystem":{"command":"npx","args":["-y","@modelcontextprotocol/server-filesystem","."]}}' --scope project
waterbrother mcp list

Create source-of-truth file

Waterbrother can load WATERBROTHER.md from your project root into the system prompt. This is useful for stable instructions and project constraints.

# inside interactive mode
/memory init
/memory add Always run tests before final answer.
/memory reload