Skip to main content
Feedback

MCP Gateway metrics

The MCP Gateway server exposes detailed Prometheus metrics about tool calls available at: http://localhost:3000/metrics

tool_call_duration_ms histogram

This is a Prometheus histogram that tracks the duration of each tool call in milliseconds.

Labels

Each metric entry may include one or more labels for filtering and aggregation. Labels are included only if non-empty.

LabelDescription
tool-nameName of the tool being called
error"true" if the call failed, "false" if successful
agentName of the calling agent or service
llm (optional)LLM provider name (if available)
model (optional)Model ID used for the request (if available)
consumer-tag (optional)The consumer tag for the request (if provided)

Example output

tool_call_duration_ms_bucket{le=“50”,tool-name=“maps_directions”,error=“false”,agent=“mcpx”} 2
tool_call_duration_ms_count{tool-name=“maps_directions”,error=“false”,agent=“mcpx”} 2
tool_call_duration_ms_sum{tool-name=“maps_directions”,error=“false”,agent=“mcpx”} 67`

Bring your own Prometheus metrics, or use the Agent Metrics dashboard in MCP Gateway control plane interface for per-agent activity, error rates, and 30-day history.

On this Page