AI inbox watcher that auto-responds to emails using Claude. A demo product for the mails-agent ecosystem.

How it works: Email arrives at your mails address -> webhook fires -> Claude generates a reply -> reply is sent back. The Worker verifies signed webhooks when configured and skips emails it has already replied to.

1. Create a .env file:

ANTHROPIC_API_KEY=sk-ant-...
MAILS_API_URL=https://api.mails0.com
MAILS_API_TOKEN=mk_your_mailbox_api_key
MAILS_MAILBOX=you@yourdomain.com
# Optional:
# WEBHOOK_SECRET=same-secret-as-mails-worker
# SYSTEM_PROMPT="You are a sarcastic robot who replies in haiku."

2. Deploy:

npx mails-monitor

3. Set the deployed Worker URL as your mailbox webhook in mails-agent.

For self-hosted mails-agent Workers, set MAILS_API_URL to your Worker URL and use your Worker token. For the hosted https://api.mails0.com service or any mk_ token, mails-monitor automatically uses /v1/* routes.

That's it. Every inbound email will get an AI-generated reply.

Single Cloudflare Worker. Zero dependencies.

Incoming Email
     |
     v
mails-agent Worker (stores email, fires webhook)
     |
     v
mails-monitor Worker (this project)
  1. Fetches full email via the mails API
  2. Calls Claude API to generate reply
  3. Sends reply via the mails API with `in_reply_to`

npm install
npm run dev    # Local dev server via wrangler
npm run deploy # Deploy to Cloudflare

MIT