mttrly - AI-powered incident management and server monitoring via MCP.
@mttrly-hq
Remote Streamable HTTP MCP server for AI SRE workflows: server health, alerts, diagnostics, playbooks, approvals, audit history, deploy verification, and safe server inspection. Uses OAuth 2.1 with PKCE.
概览
What is mttrly?
mttrly is a remote Streamable HTTP MCP connector that provides AI-powered incident management and server monitoring. It integrates with the mttrly Central API to access an AI SRE agent for server health, diagnostics, playbooks, approvals, and audit history. It is designed for users who need to monitor and manage remote servers through MCP-compatible clients.
How to use mttrly?
Connect via the mttrly connector UI in Claude.ai or Claude Desktop by adding a custom connector with URL https://api.mttrly.com/mcp and completing OAuth. For Claude Code, run claude mcp add --transport http mttrly https://api.mttrly.com/mcp. For OpenAI Codex, add the URL to ~/.codex/config.toml and run codex mcp login mttrly. Cursor users can add the same URL as a remote MCP server. Always call mttrly_get_capabilities first to discover available tools and plan restrictions.
Key features of mttrly
- Remote Streamable HTTP transport via Central API
- OAuth 2.1 authentication with PKCE S256
- 40 tools: 21 read and 19 execute
- Server health monitoring (CPU, RAM, disk, alerts)
- Non‑destructive diagnostics and investigation bypass
- Playbook execution with optional approval
- Audit logs with source attribution
Use cases of mttrly
- Monitor server health and receive proactive alerts
- Run remote diagnostics and triage incidents
- Execute remediation playbooks with approval workflows
- Investigate server issues with session‑wide bypass
- Retrieve audit trails for compliance and root‑cause analysis
FAQ from mttrly
What transport does mttrly use?
mttrly uses remote Streamable HTTP; it does not expose a local stdio surface.
How do I authenticate with mttrly?
Authentication uses OAuth 2.1 with PKCE S256 against https://app.mttrly.com. Browser-origin clients must be allowlisted.
What are the plan levels?
Two plans exist: Watchdog (read‑only tools, e.g., server status, alerts) and Deployment Bro (full read/execute tools, including diagnostics and playbooks).
Are all tools immediately available?
No. Always call mttrly_get_capabilities first to see available tools, restricted tools, and plan limits.
What happens when an action requires approval?
Actions such as playbook execution or shell commands return pending_approval. Pending actions expire after 30 minutes; use mttrly_approve_action or mttrly_cancel_action after explicit user confirmation.
