Hi Hugging Face community — we built an open-source stdio MCP server that lets AI agents work with email through EngageLab Email.
The goal is to make email a practical tool for agent workflows, not just a one-off API call. An agent can send a message, inspect incoming mail, reply in a thread, and use email as part of a larger business workflow.
What it supports today:
- Send email with plain text or HTML content
- Receive and inspect messages
- Reply to an existing message/thread
- Run locally over the standard MCP stdio transport
- Use the same email capability from the EngageLab Email CLI
Quick start:
npx -y @engagelabemail/mcp
Set the API credential in your local environment:
export ENGAGELAB_EMAIL_SECRET_KEY="your-secret-key"
Source code:
For developers who want to try the service:
A few example agent workflows:
- An internal invoice assistant that submits an approval request and notifies the right reviewers
- An e-commerce support agent that replies to a customer and keeps the conversation in one thread
- A scheduled agent that sends a daily operational summary to a team mailbox
- A coding agent that sends a test report or deployment notification
Current limitations:
- This is a local stdio MCP server; we have not added a hosted Remote MCP endpoint yet
- Mailbox and sender setup are currently managed in the EngageLab Email console
- Custom-domain review may require additional time
- We are starting with polling-based inbox inspection rather than a full event-driven inbox experience
We built this as an EngageLab Email team project and would value feedback from the community:
- Which agent email workflows are most useful to you?
- Is stdio the right first integration surface, or would a hosted MCP transport be more valuable?
- What would make this easier to try in a Hugging Face Space or another agent platform?
Thanks for taking a look — practical integration feedback is very welcome.