MCPJungle uses the OpenTelemetry SDK to emit metrics and exposes them in the Prometheus text format at the /metrics endpoint. The metrics are well-suited for scraping with Prometheus and visualising in Grafana.
Enabling metrics
In enterprise mode, OpenTelemetry is enabled automatically — no extra configuration is required.
In development mode, metrics are disabled by default. To enable them without switching to enterprise mode, set the OTEL_ENABLED environment variable before starting the server:
export OTEL_ENABLED=true
# Optionally attach resource attributes to every metric (e.g., deployment environment)
export OTEL_RESOURCE_ATTRIBUTES=deployment.environment.name=enterprise
mcpjungle start
OTEL_ENABLED accepts true / 1 to enable and false / 0 to disable. In enterprise mode you can use OTEL_ENABLED=false to opt out of metrics collection.
Accessing the metrics endpoint
Once the server is running with metrics enabled, scrape the endpoint:
curl http://localhost:8080/metrics
# HELP mcpjungle_tool_calls_total Total number of tool calls
# TYPE mcpjungle_tool_calls_total counter
mcpjungle_tool_calls_total{mcp_server_name="calculator",tool_name="calculator__add",outcome="success"} 42
mcpjungle_tool_calls_total{mcp_server_name="github",tool_name="github__git_commit",outcome="error"} 3
# HELP mcpjungle_tool_call_latency_seconds Latency of tool calls in seconds
# TYPE mcpjungle_tool_call_latency_seconds histogram
mcpjungle_tool_call_latency_seconds_bucket{mcp_server_name="calculator",tool_name="calculator__add",outcome="success",le="0.1"} 40
mcpjungle_tool_call_latency_seconds_bucket{mcp_server_name="calculator",tool_name="calculator__add",outcome="success",le="+Inf"} 42
mcpjungle_tool_call_latency_seconds_sum{mcp_server_name="calculator",tool_name="calculator__add",outcome="success"} 1.243
mcpjungle_tool_call_latency_seconds_count{mcp_server_name="calculator",tool_name="calculator__add",outcome="success"} 42
Available metrics
| Metric | Type | Description |
|---|
mcpjungle_tool_calls_total | Counter | Total number of tool calls, including prompt calls. |
mcpjungle_tool_call_latency_seconds | Histogram | Latency in seconds for each tool call. |
| Label | Description |
|---|
mcp_server_name | Name of the upstream MCP server that owns the tool. |
tool_name | Canonical tool name (server__tool, e.g. calculator__add). |
outcome | success or error. |
0.005, 0.01, 0.025, 0.05, 0.1, 0.25, 0.5, 1, 2, 5, 10, 20, 30 seconds.
Using Prometheus and Grafana
The production Docker Compose file (docker-compose.prod.yaml) ships with a Prometheus container pre-configured to scrape MCPJungle every 15 seconds:
scrape_configs:
- job_name: 'mcpjungle'
static_configs:
- targets: ['mcpjungle:8080']
metrics_path: '/metrics'
scrape_interval: 15s
curl -O https://raw.githubusercontent.com/mcpjungle/MCPJungle/refs/heads/main/docker-compose.prod.yaml
docker compose -f docker-compose.prod.yaml up -d
http://localhost:9090. From there you can connect a Grafana instance (not included in the compose file) and build dashboards using the mcpjungle_* metrics.
Use OTEL_RESOURCE_ATTRIBUTES to attach additional metadata (such as cluster name or deployment environment) to all metrics emitted by a specific MCPJungle instance. This is useful when you run more than one instance and want to filter metrics by environment in Grafana.
