Simple Observability MCP

The Model Context Protocol (MCP) allows your AI agents to query real-time monitoring data. By connecting your agent to our MCP server, it can fetch metrics, logs, active incidents, and other infrastructure details directly during your conversation.

Configuration

Add the following to your MCP client configuration:

{
  "mcpServers": {
    "simob": {
      "url": "https://mcp.simpleobservability.com/sse",
      "headers": {
        "SIMOB-API-KEY": "YOUR_API_KEY"
      }
    }
  }
}

Authentication

To authenticate the connection, you need a valid API key. Generate one from your dashboard by navigating to Settings > API keys. The keys page also shows the MCP configuration with your key pre-filled, so you can copy and paste directly.

Treat this key like a password. It grants your agent read access to your infrastructure data.

Available tools

The MCP server exposes six tools. Your agent picks the right one based on your conversation, so you do not need to call them manually.

list_open_incidents lists currently open incidents. Each incident is a failure detected by a server check, job execution, alert rule evaluation, or endpoint monitor.

get_incident_context returns a fully detailed view of a single incident, including the attached entity (server, job, alert rule, or endpoint). This collapses what would otherwise be multiple separate calls into one.

search_entities searches across the platform by name. It covers servers, alert rules, endpoints, jobs, services, and notification channels. Use it when you need to find an entity by name and get its ID.

get_available_sources lists all metric names and log source names available in the platform. Call this before querying metrics or logs, since you need the exact source name to build a query.

query_metrics queries metric data over a time range or at the current instant. Pass a metric name and optional selectors to filter by labels.

query_logs queries log entries over a time range. You can filter by source, apply a transformation pipeline, and aggregate results.

Usage examples

Here are a few prompts you can try once your agent is connected:

• “I have an open incident, look into it.” The agent will list open incidents, fetch the full context for the relevant one, and investigate using metrics and logs.
• “Search for the production web server and show me its CPU usage.” The agent will search for the server by name, discover available metrics, and query the CPU metric.
• “Show me the last 30 minutes of nginx logs.” The agent will discover the nginx log source and query the logs.
• “What endpoints are currently down?” The agent will search for endpoints, check their status, and report back.

Privacy and security

The MCP server uses your API key to authenticate every request. The key is sent as a header (SIMOB-API-KEY) and is never stored by the MCP server itself. All requests are read-only, so the agent cannot modify your infrastructure.