MCP (Model Context Protocol)

GlitchTip includes a built-in MCP server that allows AI assistants to interact with your error tracking, performance, and log data.

Enabling MCP

Set the following environment variable on your GlitchTip server:

GLITCHTIP_ENABLE_MCP=True

The MCP endpoint is available at:

https://your-glitchtip.example.com/mcp

The MCP server uses OAuth 2.0 for authentication. AI clients authenticate using GlitchTip API tokens. The server supports dynamic client registration, so compatible MCP clients can register automatically.

Connecting an AI Client

Claude Desktop

Add GlitchTip as an MCP server in your Claude Desktop configuration:

{
  "mcpServers": {
    "glitchtip": {
      "url": "https://your-glitchtip.example.com/mcp"
    }
  }
}

Claude Desktop will handle the OAuth flow automatically when you first connect.

Other MCP Clients

Any MCP-compatible client can connect to the endpoint. The server uses the Streamable HTTP transport and advertises its OAuth configuration for automatic discovery.

Available Tools

The MCP server exposes 17 tools across several categories:

Organizations & Projects

list_organizations — List all organizations the authenticated user can access
list_projects — List all projects in an organization

Issue Management

list_issues — List issues, optionally filtered by project. Use query='is:unresolved' for active issues
get_issue — Get details for a single issue by ID
get_latest_event — Get the latest event for an issue
get_event — Look up a specific event by ID, returned with its parent issue
update_issue — Resolve, unresolve, or ignore an issue

Performance Monitoring

list_transaction_groups — List endpoints/operations with performance stats (avg duration, p50, p95)
get_transaction_group — Get details for a single transaction group
list_transaction_spans — Get span breakdown for a transaction group (requires DuckDB)
list_span_groups — Query span groups across the organization (e.g., slow DB queries). Use op='db' to find slow database queries (requires DuckDB)
detect_n_plus_one — Detect N+1 query patterns across transactions (requires DuckDB)
get_transaction_trend — Get daily performance trend for a transaction group (requires DuckDB)

Alerting & Monitoring

list_alerts — List alert rules for an organization, optionally filtered by project
list_monitors — List uptime monitors for an organization

Logs (when `GLITCHTIP_ENABLE_LOGS=True`)

list_logs — Search log events by level, service, environment, trace ID, or text query
get_log — Get a single log event by ID

Use Cases

With the MCP server, AI assistants can:

Investigate errors — Browse issues, inspect stack traces and event data, and resolve issues
Analyze performance — Find slow endpoints, detect N+1 query patterns, and identify performance regressions
Query logs — Search application logs and correlate them with traces
Monitor status — Check alert configurations and uptime monitor status