Email, built for AI agents.
Most platforms bolt a chatbot onto a dashboard. Drin does the opposite: everything email can do is exposed as agent-native tools — a 54-tool MCP server, a full CLI, Claude Skills, and an inbox an agent can reply from.
Four ways to put email in an agent's hands.
MCP server
The whole API as typed tools — send, batch, domains, keys, templates, webhooks, contacts, inbox and reply. Works with any MCP client.
54 toolsCommand-line
A complete CLI that mirrors the API — scriptable sends, domain verification, inbox triage and replies from any shell or CI step.
Full surfaceClaude Skills
Drop-in skills that teach Claude how to use Drin — composing, triaging an inbox, verifying a domain and reading metrics, end to end.
4 skillsAgent inbox
Give an agent its own address. Inbound arrives as structured events, it reads the thread, and replies in-thread — not just fire-and-forget sends.
Send + receiveBring the client you already use.
Drin speaks the Model Context Protocol, so any MCP client connects to the same 54 tools — Claude, Cursor, Codex and more. Vibe-coding on Lovable or Replit? Use the plain REST API. Pick a surface to wire it up.
Every client, identical behaviour. However your agent calls it, the result is the same.
An agent that can only send is half an agent. The interesting work happens when it can receive and reply.
Hand an agent a send endpoint and it can fire off a notification. Give it its own address, structured inbound events, and a reply tool, and it can hold a conversation — confirm an order, answer a support thread, chase a missing detail, escalate when it's stuck.
That is the wedge. Drin was built inbox-first, so the same primitives that power human mail — threads, replies, suppression, deliverability — are the ones your agent gets. No marketing-automation cruft, no "AI credits": just email, exposed as tools.
Give your agent an inbox today.
Grab a key, point your MCP client at Drin, and let an agent send, receive and reply. Test in the sandbox; production starts at $3 a month.