MCP server
Messy speaks the Model Context Protocol, so an AI agent can connect and work your account directly: send messages, manage templates and segments, run campaigns and drips, configure channels, work the inbox and read analytics. Agents connect through a real OAuth sign-in, act only in the environment you approve and at the connecting user's own permissions, and every tool call is logged. Turn the whole thing off, disable a teammate, or revoke a single agent from one screen.
How an agent connects
Add connector
Paste the MCP URL
Authorize
OAuth sign-in + consent
Call tools
The whole API surface
Audited
Every call is logged
OAuth, not a shared key
Agents connect with an OAuth 2.1 sign-in and PKCE, then you approve which environment each one may touch.
The whole surface, as tools
Everything the API can do is exposed as MCP tools, scoped to the connecting user's role. Admin-only tools stay hidden from members.
Governed and audited
A master switch, per-user and per-connection controls, and a usage log of every call an agent makes.