Model Context Protocol (MCP)
An open protocol that lets AI models call external tools and data sources through a standard interface — so capabilities show up as ready-made tools in any MCP client.
The Model Context Protocol (MCP) is an open standard for connecting AI models to external tools and data. An MCP server exposes a set of typed tools; an MCP client — Claude, Cursor, and others — discovers those tools and lets the model call them.
It matters because it replaces bespoke, per-integration glue with one consistent interface. A capability published over MCP works in any MCP client without custom wiring.
How Mailgent uses MCP
Mailgent runs an MCP server that turns each capability into tools: mail_send, vault_get, vault_totp, calendar_create, identity_sign, and more. Point a client at it with the agent's key and the tools appear, scope-gated to what that agent may do.
You can run the server locally with npx, or use the hosted endpoint at api.mailgent.dev/mcp.
{ "mcpServers": { "mailgent": {
"command": "npx",
"args": ["-y", "@mailgent-dev/mcp"],
"env": { "LOOMAL_API_KEY": "mgnt-..." } } } }Tools, not endpoints
Because tools are typed and self-describing, the model knows what arguments each takes and what it returns. That makes capabilities like sending email or paying an invoice safe to expose directly to an agent, gated by scopes.
FAQ
Which clients support MCP?
Claude, Cursor, and a growing set of clients and frameworks. Mailgent's server works with any of them.
Do I have to use MCP?
No. Every capability is also available over REST. MCP is the fastest path for tool-using agents.
Give your agent an inbox.
A real email address, a vault, 2FA, and an identity in one API call.