I built email infrastructure for AI agents: send, receive, parse, and reply over MCP

Most email APIs are built around one direction: your app sends a transactional email and a human reads it somewhere else.

That breaks down when the user is an AI agent.

Agents often need the full loop:

• send email
• receive replies
• parse inbound messages into structured data
• keep conversation/thread context
• reply through an API, CLI, or MCP tool
• trigger webhooks back into the app
• require approval before risky outbound sends

So we built Drin: agent-native email infrastructure for developers and AI agents.

The goal is to give an app or agent a real inbox, not just a send endpoint.

A few examples of workflows we are designing for:

• a support agent that can receive a customer reply, inspect thread context, draft a response, and wait for approval
• an AI SDR that can read inbound replies but cannot send externally without an auditable gate
• a devtool or CI agent that needs a durable email identity for signups, receipts, alerts, and replies
• an app that needs inbound parsing, reply APIs, and webhooks without stitching together multiple providers

The product supports send, receive, inbound parsing, threads, replies, webhooks, CLI, MCP, sandbox testing, and production domains.

Free sandbox is live. Production starts at $3/month.

Site: https://drin.run/
Docs: https://docs.drin.run/

I am looking for feedback from people building AI agents, support tools, CRMs, devtools, or anything that needs programmable inbound email. If you have tried to build this with Resend, SendGrid, Mailgun, SES, Gmail, or Postmark, I would especially like to hear what broke first.