This dashboard provides operational visibility into the SigNoz MCP server. It tracks tool-call volume and error rates, MCP (JSON-RPC) method usage, and execute_tool
span latency (broken down per tool) so teams can quickly see how the server is being used, which tools are slow or failing, and whether the overall service is healthy.
Dashboard Preview
Dashboard JSON
Recommended. Uses the [V2 dashboard schema](https://signoz.io/docs/dashboards/dashboards-v2-api/) and needs SigNoz v0.135.0 or newer.
Import it in SigNoz with **Dashboards → + New dashboard → Import JSON**. [Import guide](https://signoz.io/docs/dashboards/import-dashboard/)
What This Dashboard Monitors
This dashboard tracks critical operational metrics for your SigNoz MCP server to help you:
Track Usage: Monitor total tool-call volume to get a quick pulse on overall traffic and adoption of the MCP server.** Ensure Reliability**: Watch the MCP span error rate to get an at-a-glance signal of service health and catch sudden spikes early.** Understand the Traffic Mix**: See per-tool call rates to identify which tools drive the most load and how usage is distributed.** Pinpoint Failures**: Break down tool calls and error rates by tool to find exactly which tool is failing rather than just the aggregate.** Optimize Performance**: Trackexecute_tool
span latency percentiles (p50/p95/p99), including per-tool p95, to surface both typical and tail latency.Analyze Protocol Usage: Observe per-method JSON-RPC call rates to understand how clients interact with the MCP protocol.
Metrics Included
Overview
Total Tool Calls: A single-value view of the total number of tool calls handled by the MCP server over the selected time range. It's a quick pulse on overall traffic and adoption.MCP Spans Error Rate: A single-value summary of the percentage of MCP spans that resulted in an error. It gives an at-a-glance signal of service health. Low and stable is good, sudden spikes warrant investigation.Total Tool Calls (by tool): A pie chart breakdown of total tool calls by tool. It complements the rate graphs with a clear share-of-usage view across the full time range.
Tool Usage
Tool calls / sec by tool: Per-tool call rate, stacked to show the traffic mix across all tools. This reveals which tools are driving load and how usage is distributed.Tool calls by error flag: Tool calls split bymcp.tool.is_error
(true = errored). This separates healthy calls from failures, making it easy to spot when errors start climbing relative to total volume.MCP method calls / sec: Per-method JSON-RPC call rate (initialize
,ping
,tools/list
,resources/list
,prompts/list
, and more). This shows how clients are interacting with the protocol and which methods dominate traffic.
Performance & Reliability
execute_tool span latency: p50 / p95 / p99 ofexecute_tool
span duration. Excludes the SSE long-pollGET /mcp
. Tracking these percentiles surfaces both typical performance and tail latency that affects the worst-case user experience.execute_tool p95 latency by tool: p95 ofexecute_tool
span duration broken down by tool (gen_ai.tool.name
). This identifies exactly which tool is slow, so optimization effort can be targeted.execute_tool error rate % by tool: Per-tool error rate %, computed asexecute_tool
spans withhas_error=true
over total, grouped by tool (gen_ai.tool.name
). This pinpoints which specific tools are failing rather than just the aggregate rate.