Tools that match how agents work
List inboxes, read threads and messages, send, reply and forward, add and remove labels, create and schedule drafts, read events, manage calendar events, extract verification codes and see held messages. The tools return the reply text without quoted history, so context stays small.
Deliberately no approve tool
Approving a held message is a human action in Innmail. The MCP server has no tool for it, and neither does any API key. An agent can see that its message is waiting, explain why, and move on; a person approves in the Approvals page.
Scoped keys
Create a key for one agent, one pod or the workspace. An agent-scoped key only sees its own inbox, so one MCP configuration per agent keeps tenants and customers apart.
Not only MCP
The same inbox is reachable over REST with an OpenAPI 3.1 spec, signed webhooks, an SSE stream and SDKs for TypeScript, Python and Go. Mix them: let the agent act through MCP while your backend listens to webhooks.
Frequently asked questions
Which clients work?
Any MCP client that supports Streamable HTTP with a Bearer token, including Claude Desktop and Claude Code. Setup snippets are on the Developers page after sign-in.
Is the MCP server hosted?
Yes. It runs on www.innmail.app; there is nothing to install or run locally.
Do I need to connect Gmail?
No. Agent inboxes live on @innmail.app or on your own verified domain. Google sign-in shares only your name and email address; Innmail never asks for mailbox access.
What does it cost?
Nothing during the private beta: 3 @innmail.app inboxes and 100 outside recipients a day per workspace, unlimited inboxes on a domain you verify. Mail between your own agents is not counted.
Give every AI agent its own inbox.
A real @innmail.app address in one API call, or one on your own domain. Free in the private beta, no Gmail to connect.
Request beta access