/api/mcp. Connect an AI agent to it and the agent can search logs, count them over time, list top field values, find slow or failing spans, and read whole traces. It cannot change anything.
Before you start
Create a personal API key so the agent acts as you. For a shared agent, use a service account key instead.Claude Code
Cursor
Add the server to~/.cursor/mcp.json, or to .cursor/mcp.json in a project:
Claude Desktop and other stdio clients
Clients that only start local processes can reach the server throughmcp-remote. For Claude Desktop, add this to claude_desktop_config.json:
Tools
Times accept ISO 8601 with a timezone, such as
2026-10-06T14:30:00Z, or relative values such as now, now-15m, and now-7d. Results are trimmed to fit an agent’s context: long values are cut, large traces return their first 200 spans, and a result too large for the agent comes back as an error asking for a narrower query.
Try it
Ask your agent: “Which services logged the most errors in the last hour, and when did it start?”Access and limits
- The server accepts only a bearer key, never a browser session.
- Agent requests share the key owner’s read rate limit (
READ_RATE_LIMIT) with their searches in the UI. - Log searches, span searches, and trace lookups appear to admins in Settings → Activity as API key searches.

