MCP server

Connect AI agents and assistants to Maetra through the Model Context Protocol. One authenticated endpoint exposes up to 18 capability-based tools for Task Guard, Govern, and Secure.

Use the MCP server to keep work tied to the user's task, scan prompts and tool calls, request approval before consequential actions, authorize the exact provider request once, and preserve execution attempts and observed effects. The tool list follows the workspace settings and API-key scopes, so a client only sees the controls it is allowed to use.

Start with Overview & connection, then run the end-to-end MCP example.

Overview & connectionConnect an MCP client to Maetra, authenticate requests, and understand the supported transport and protocol.Tools referenceReference capability access, Task Guard, content scanning, approvals, one-use execution authorization, attempts, effects, and active policies.End-to-end exampleInitialize an MCP session, request approval, authorize one exact execution, record provider attempts and effects, and complete the task.