openwork/packages/owpenbot

Owpenbot

Simple Slack + WhatsApp + Telegram bridge for a running OpenCode server.

Install + Run (WhatsApp)

One-command install (recommended):

curl -fsSL https://raw.githubusercontent.com/different-ai/openwork/dev/packages/owpenbot/install.sh | bash

Or install from npm:

npm install -g owpenwork

Quick run without install:

npx owpenwork

Then follow the guided setup (choose what to configure, link WhatsApp, start).

1. One-command setup (installs deps, builds, creates .env if missing):

pnpm -C packages/owpenbot setup

2. (Optional) Fill in packages/owpenbot/.env (see .env.example).

Required:

• OPENCODE_URL
• OPENCODE_DIRECTORY
• WHATSAPP_AUTH_DIR

Recommended:

• OPENCODE_SERVER_USERNAME
• OPENCODE_SERVER_PASSWORD

3. Run owpenwork and follow the guided setup:

owpenwork

Owpenwork keeps the WhatsApp session alive once connected.

6. Pair a user with the bot (only if DM policy is pairing):

• Run owpenwork pairing list to view pending codes.
• Approve a code: owpenwork pairing approve <code>.
• The user can then message again to receive OpenCode replies.

Usage Flows

One-person flow (personal testing)

Use your own WhatsApp account as the bot and test from a second number you control.

1. Run owpenwork and choose “personal number.”
2. Scan the QR when prompted.
3. Message yourself or from a second number; your number is already allowlisted.

Note: WhatsApp’s “message yourself” thread is not reliable for bot testing.

Two-person flow (dedicated bot)

Use a separate WhatsApp number as the bot account so it stays independent from your personal chat history.

1. Create a new WhatsApp account for the dedicated number.
2. Run owpenwork and choose “dedicated number.”
3. Scan the QR when prompted.
4. If DM policy is pairing, approve codes with owpenwork pairing approve <code>.

Telegram (Untested)

Telegram support is wired but not E2E tested yet. To try it:

• Run owpenwork login telegram --token <token>.
• Or set TELEGRAM_BOT_TOKEN and TELEGRAM_ENABLED=true.

Slack (Socket Mode)

Slack support uses Socket Mode and replies in threads when @mentioned in channels.

1. Create a Slack app.
2. Enable Socket Mode and generate an app token (xapp-...).
3. Add bot token scopes:
   • chat:write
   • app_mentions:read
   • im:history
4. Subscribe to events (bot events):
   • app_mention
   • message.im
5. Set env vars (or save via owpenbot slack set-tokens ...):
   • SLACK_BOT_TOKEN=xoxb-...
   • SLACK_APP_TOKEN=xapp-...
   • SLACK_ENABLED=true

Commands

owpenwork
owpenwork --non-interactive
owpenwork login whatsapp
  owpenwork login telegram --token <token>
  owpenwork slack status
  owpenwork slack set-tokens <xoxb> <xapp>
  owpenwork pairing list
  owpenwork pairing approve <code>
  owpenwork status
  owpenwork doctor --reset

Defaults

• SQLite at ~/.openwork/owpenbot/owpenbot.db unless overridden.
• Config stored at ~/.openwork/owpenbot/owpenbot.json (created by owpenwork or owpenwork setup).
• DM policy defaults to pairing unless changed in setup.
• Group chats are disabled unless GROUPS_ENABLED=true.

Tests

pnpm -C packages/owpenbot test:unit
pnpm -C packages/owpenbot test:smoke
pnpm -C packages/owpenbot test:cli
pnpm -C packages/owpenbot test:npx