MCP Integration

MiniAPM will include a built-in Model Context Protocol (MCP) server that allows AI assistants like Claude to query your application’s performance data and errors.

What is MCP?

The Model Context Protocol is a standard for connecting AI models to external data sources and tools. With MCP, you’ll be able to ask Claude questions like:

“What errors happened in production today?”
“Which routes are slowest?”
“Show me details about this error”
“What’s the system status?”

Planned Tools

Read-only

Tool	Description
`list_errors`	Find recent errors with filtering
`get_error`	Get full error information with stack traces
`error_trends`	Hourly error counts
`list_traces`	Find traces with filtering
`get_trace`	Full span waterfall with N+1 detection
`slow_traces`	Find traces above a duration threshold
`route_summary`	P50/P95/P99, throughput, error rate per route
`list_deploys`	Recent deployments
`deploy_impact`	Before/after comparison for a deploy
`system_status`	Overall health overview
`detect_n_plus_one`	N+1 patterns in a specific trace

Write

Tool	Description
`resolve_error`	Mark error as resolved
`ignore_error`	Mark error as ignored
`reopen_error`	Reopen a resolved/ignored error

Deployment Modes

Two modes are planned:

Stdio Mode (Local)

For local development with Claude Desktop:

miniapm mcp

HTTP Mode (Remote)

For remote or shared access, the MCP endpoint will be available at POST /mcp with API key authentication.