Farclaw

Open-source OpenClaw template for running a 5-agent Farcaster content team with Telegram control.

What you get

• Opinionated multi-agent setup:
  • orchestrator (routes work, publishes after approval)
  • strategist (content planning)
  • drafter (draft variants)
  • tone (tone bootstrap and adaptation)
  • engager (reply opportunity scouting)
• Telegram inbound routing to orchestrator
• Neynar skill for Farcaster read/write workflows
• Web scraping skill with JS fallback
• Docker + Railway-friendly deployment

Quick start (local)

1. Copy env file:

cp .env.example .env

2. Fill required env vars:

• OPENROUTER_API_KEY
• TELEGRAM_BOT_TOKEN
• NEYNAR_API_KEY
• NEYNAR_SIGNER_UUID

3. Optional but recommended:

• PROJECT_NAME
• FARCASTER_HANDLE
• FARCASTER_FID
• FARCASTER_CHANNEL_ID (optional default channel slug)
• TELEGRAM_ALLOWED_USER_ID (restrict who can DM the bot)

4. Start:

docker compose up --build

5. Pair your Telegram user (if needed):

openclaw pairing list telegram
openclaw pairing approve telegram <CODE>

Environment variables

See .env.example for full list. Key customization points:

• PROJECT_NAME: shown in agent naming and scraper user-agent
• FARCASTER_FID: source account used for one-off tone bootstrap
• FARCASTER_CHANNEL_ID: default channel slug used for publish/reply when set
• PRIMARY_MODEL / FALLBACK_MODEL: OpenRouter model selection
• RAILWAY_DOMAIN: optional control UI CORS origin (without scheme)

Customizing project knowledge

Edit workspaces/orchestrator/KNOWLEDGE.md with your domain facts and messaging.

At startup, this file is synced to all agent workspaces so every agent shares the same context.

Tone profile bootstrap

Ask the orchestrator:

"Bootstrap tone profile from my cast history (FID )."

The tone agent should fetch historical casts once and write guidance to TONE_PROFILE.md.

Railway deploy (step-by-step)

1. Click the deploy button above (or use the direct link) and create a new Railway project.
2. Add a persistent volume and mount it at /data.
3. Confirm service port is 8080 and set:
   • PORT=8080
4. Fill required provider secrets:
   • OPENROUTER_API_KEY from OpenRouter Keys: https://openrouter.ai/keys
   • TELEGRAM_BOT_TOKEN from BotFather: https://t.me/BotFather
   • BotFather setup guide: https://core.telegram.org/bots#6-botfather
   • NEYNAR_API_KEY from Neynar dashboard: https://neynar.com/
   • NEYNAR_SIGNER_UUID (create signer + approve once): https://docs.neynar.com/docs/write-to-farcaster-with-signer
   • Easy option in Neynar dashboard: Apps -> Your App -> Agents tab -> Use existing account (copy signer UUID)
5. Set optional but recommended Farcaster identity fields:
   • FARCASTER_HANDLE (without @)
   • FARCASTER_FID (your numeric FID)
   • FARCASTER_CHANNEL_ID (default channel slug)
6. Set security/config variables:
   • SETUP_PASSWORD=${secret(32)}
   • OPENCLAW_GATEWAY_TOKEN=${secret(64)}
   • These values are editable, but with ${secret(...)} they are auto-generated by Railway at deploy time, so you usually do not need to change them manually.
   • TELEGRAM_ALLOWED_USER_ID=<your telegram numeric user id> (recommended)
7. Add RAILWAY_DOMAIN to your generated Railway domain (no https://).
8. Deploy, then open logs and confirm boot finished without missing env errors.
9. Pair Telegram only if you did not set TELEGRAM_ALLOWED_USER_ID:

openclaw pairing list telegram
openclaw pairing approve telegram <CODE>

10. Send a DM to your bot and bootstrap tone profile first:

• ask: Bootstrap tone profile from my cast history (FID <your fid>).
• then test content planning
• then test draft generation
• then test publish approval flow

11. Customize your project context before real usage:

• edit workspaces/orchestrator/KNOWLEDGE.md with your product, audience, voice, and constraints
• this file is synced to all agent workspaces at startup

Safety defaults

• Orchestrator is the only agent with write actions.
• Publishing should happen only after explicit approval.
• Tone history fetch is one-off bootstrap, not per-request.