OAuth 2.140 toolsApproval gatesAudit log

Incident Management MCP Server

Alerts, diagnostics, playbooks — from Claude Code, Cursor, or Codex

Connect mttrly to your AI coding assistant through Model Context Protocol. View alerts, run scoped diagnostics, request approval-gated remediation playbooks, and review the audit trail without exposing raw SSH as the normal AI control path.

Endpoint:https://api.mttrly.com/mcp

Comparing monitoring MCP servers with remediation tools? Read about the incident response action layer for MCP.

Create account with email See MCP setup

Required before MCP setup

Account → server → agent → OAuth

The MCP endpoint is not an anonymous connection to arbitrary servers. It exposes the servers already linked to your hosted mttrly account through the required outbound agent.

Create account

Add server

Start onboarding and create the server-specific install token.

Install agent

Run the token-specific systemd installer shown in your account.

Connect MCP

Add the endpoint to your client and complete OAuth in the browser.

Want the architecture and safety boundaries first? Read how mttrly manages a VPS without raw SSH and how it works as an incident response action layer.

Quick Start

Create an account, connect a server through the required agent, then add the mttrly endpoint to your MCP client and complete OAuth.

Claude Code

bash

claude mcp add mttrly --transport http https://api.mttrly.com/mcp

Run this in your terminal. OAuth login opens in browser automatically.

Cursor

json

{
  "mcpServers": {
    "mttrly": {
      "url": "https://api.mttrly.com/mcp"
    }
  }
}

Go to Cursor → Settings → Cursor Settings → MCP → Add server.

Claude Desktop

connector

Settings -> Connectors -> Add custom connector -> https://api.mttrly.com/mcp

Use the Connectors UI in Claude Desktop or Claude.ai.

OpenAI Codex

toml

[mcp_servers.mttrly]
url = "https://api.mttrly.com/mcp"

Add to ~/.codex/config.toml. Run codex mcp login mttrly for OAuth.

What you can ask

Talk to your servers in plain English. Here are some examples.

> Show me all critical alerts across my servers

Returns active alerts with severity, timestamps, and affected servers. Watchdog plan and above.

> Run a diagnostic on prod-web-01 — it feels slow

AI-powered diagnostic checks CPU, RAM, disk, processes, and logs. Returns summary with raw output. 30s cooldown.

> What playbooks are available for nginx issues?

Lists remediation playbooks filtered by category, with required parameters and approval requirements.

> Execute the restart-nginx playbook on prod-web-01

Creates a pending action. You confirm before anything runs. Approval and execution are recorded in the audit trail.

Available tools

Highlighted tools from the 40-tool MCP surface, organized into read and execute groups.

Watchdog highlights

mttrly_get_capabilitiesGet current plan, available tools, and limits. Call this first.

mttrly_list_serversList all connected servers with online/offline status and server IDs.

mttrly_get_server_statusDetailed server health: CPU, RAM, disk, active alerts, OS info.

mttrly_get_alertsGet alerts with severity/status filtering. Plan-based retention limits.

mttrly_list_playbooksList remediation playbooks with categories and approval requirements.

mttrly_quick_triageSafe Watchdog triage from current status and active alerts.

Execute (Deployment Bro+)

mttrly_run_playbookExecute a remediation playbook. Dangerous ones require explicit approval.

mttrly_execute_commandRequest an approval-gated command execution action. Prefer playbooks when available. Approval or bypass state and execution are recorded in the audit trail.

mttrly_get_pending_actionsList actions awaiting your approval with expiration times.

mttrly_approve_actionApprove or reject a pending action. Blocks until execution completes.

mttrly_get_audit_logAudit trail: approval decisions, command executions, playbook runs, agent fixes.

Security & Safety

OAuth 2.1

Full spec: PKCE S256, JWT RS256, refresh token rotation with reuse detection. OAuth browser login is recommended, and API keys are also supported from your dashboard.

Approval gates

Approval-required actions normally create pending actions. A user-authorized, bounded mttrly_execute_command Investigation or narrowly configured preauthorization are explicit exceptions; both remain audited.

Audit log

Approval decisions, command executions, playbook runs, and remediation actions are recorded with timestamps for review.

Plan-based access

Read tools available on all plans. Execution tools require Deployment Bro. No silent 403s — the capabilities tool tells the agent what's available.

Need to make the decision away from your IDE? Supported pending actions can be approved or rejected through Telegram.

MCP access by plan

Watchdog

Free

Starter MCP tools

Starter tools — capabilities, servers, status, alerts, quick triage, playbooks, recovery token, offline delete

Deployment Bro

$39/mo

Full access

40 tools — everything above + diagnostics, playbooks, commands, approvals, investigations, and recovery

Ready to manage servers from your IDE?

Free tier includes starter MCP access. No credit card required.

Create account and connect a server

Install in Claude Code:

claude mcp add mttrly --transport http https://api.mttrly.com/mcp

View pricingorRead docs

Need an incident to test? Start with the Linux incident playbooks.