> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mcpjungle.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Observability with OpenTelemetry

> Use OpenTelemetry to monitor mcpjungle and all MCP interactions in your infrastructure.

Mcpjungle emits Prometheus-compatible metrics using OpenTelemetry and exposes them at the `/metrics` endpoint.

This gives you a simple operational view of tool traffic and latency across your registered MCP servers.

## When metrics are enabled

* In **enterprise mode**, metrics are enabled by default.
* In **development mode**, metrics are disabled by default.

To enable them in development mode:

```bash theme={null}
export OTEL_ENABLED=true
export OTEL_RESOURCE_ATTRIBUTES=deployment.environment.name=development

mcpjungle start
```

<Note>
  You can also set `OTEL_ENABLED=false` in enterprise mode if you explicitly want to disable metrics, but that is not the recommended operating posture for shared deployments.
</Note>

## Metrics endpoint

Once Mcpjungle is running with telemetry enabled:

```bash theme={null}
curl http://localhost:8080/metrics
```

The response uses standard Prometheus text exposition format.

## Available metrics

| Metric                                | Type      | Description                                      |
| ------------------------------------- | --------- | ------------------------------------------------ |
| `mcpjungle_tool_calls_ratio_total`    | Counter   | Total number of tool calls handled by Mcpjungle. |
| `mcpjungle_tool_call_latency_seconds` | Histogram | Latency for tool calls, in seconds.              |

Both metrics include these labels:

| Label             | Description                                         |
| ----------------- | --------------------------------------------------- |
| `mcp_server_name` | Name of the upstream MCP server that owns the tool. |
| `tool_name`       | Canonical tool name, for example `calculator__add`. |
| `outcome`         | `success` or `error`.                               |

## Example output

```text theme={null}
# HELP mcpjungle_tool_calls_ratio_total Total number of tool calls
# TYPE mcpjungle_tool_calls_ratio_total counter
mcpjungle_tool_calls_ratio_total{mcp_server_name="calculator",tool_name="add",outcome="success"} 42

# HELP mcpjungle_tool_call_latency_seconds Latency of tool calls in seconds
# TYPE mcpjungle_tool_call_latency_seconds histogram
mcpjungle_tool_call_latency_seconds_count{mcp_server_name="calculator",tool_name="add",outcome="success"} 42
```

## Prometheus

The production Compose path ships with a Prometheus service already configured to scrape Mcpjungle:

```yaml theme={null}
scrape_configs:
  - job_name: 'mcpjungle'
    static_configs:
      - targets: ['mcpjungle:8080']
    metrics_path: '/metrics'
    scrape_interval: 15s
```

If you run your own Prometheus stack, scrape the gateway's `/metrics` endpoint directly.

## Feedback

If you want to see more metrics or have feedback on the current ones, please share your thoughts in the [GitHub discussions](https://github.com/orgs/mcpjungle/discussions) or open an [Issue](https://github.com/mcpjungle/MCPJungle/issues).

## Related pages

<CardGroup cols={2}>
  <Card title="Production deployment" icon="cloud" href="/deployment/production">
    Run Mcpjungle in enterprise mode with the right operational defaults.
  </Card>

  <Card title="Database" icon="database" href="/deployment/database">
    Back your shared deployment with PostgreSQL.
  </Card>

  <Card title="Governance overview" icon="lock" href="/governance/overview">
    Understand the enterprise operating model around authentication and client access.
  </Card>

  <Card title="Reference: environment variables" icon="sliders" href="/reference/environment-variables">
    See the full environment variable reference for telemetry and deployment tuning.
  </Card>
</CardGroup>