The tools reach real machines through agents you already run, which is what separates this from a monitoring dashboard: you can ask what is wrong and then run the diagnostic. The capabilities call comes first because what you can reach depends on the plan you are on.
A remote connector for mttrly, an AI SRE service for server health, diagnostics, playbooks and audit history. Every call reaches the mttrly Central API first, which then talks to the agents on your servers over their existing channel - a standalone agent exposes no local MCP surface of its own.
- Your plan's capabilities: which tools are available, which are restricted, and the plan limits - the call the vendor asks you to make before any other
- The servers connected to your account, and the status of any one of them
- Current alerts
- A quick triage pass, and a diagnostic run against a server
- A service's actual state as the agent on the machine reports it
- The playbooks available to run
An mttrly account with agents already connected, and OAuth with PKCE. The endpoint is remote, so add it as a custom connector rather than in a local stdio config. Browser-origin clients have to be allowlisted before they are accepted.
One command — claude mcp add --transport http mttrly https://api.mttrly.com/mcp