pavolloffay/opentelemetry-mcp-server

OpenTelemetry Model Context Protocol (MCP) Server

The OpenTelemetry MCP server enables LLM to efficiently use OpenTelemetry stack.

The MCP uses for the collector config validation.

https://github.com/user-attachments/assets/af56cda4-9ab0-4c93-969e-a2ee7b9f0480

Install & Run

go install github.com/pavolloffay/opentelemetry-mcp-server@latest
opentelemetry-mcp-server --protocol http --addr 0.0.0.0:8080

# or docker run --rm -it -p 8080:8080 ghcr.io/pavolloffay/opentelemetry-mcp-server:latest --protocol http --addr 0.0.0.0:8080 

claude mcp add --transport=http otel http://localhost:8080/mcp --scope user

Functionality

At the moment the MCP server offer tools to configure an OpenTelemetry collector. Tools return strict JSON schema for each collector component ensuring the configuration is correct.

A complete list of tools can be found in the .

Future work / Roadmap

• Enable LLM to understand/profile data collector is receiving.
  • What are the (resource) attributes and values?
  • It can help LLM to write PII (filtering) rules specific for a given organization.
• Enable LLM to understand which workloads are sending telemetry data.
• Enable LLM to understand how much data each workload is sending.
• Enable LLM to tweak sampling configuration based on the collected data.
• Enable LLM to size storage based on the collected data volumes.
• Enable LLM to validate OpenTelemetry transformation language.

Instrumentation

• Enable LLM to correctly instrument code.
  • Enable LLM identify uninstrumented code (e.g. RPC frameworks that should be instrumented).
• Enable LLM to identify high cardinality attributes explicitly added.
• Enable LLM to identify PII attributes explicitly added.
• Enable LLM to fix issues in the codebase based on the telemetry data.

Other OpenTelemetry MCP servers

• https://github.com/mottibec/otelcol-mcp - collector config
• https://github.com/shiftyp/otel-mcp-server - data profiling, requires OpenSearch
• https://github.com/austinlparker/otel-mcp - config, data profiling
• https://github.com/liatrio-labs/otel-instrumentation-mcp - instrumentation

https://github.com/austinlparker/otel-mcp

docker run --rm -it -v $(pwd)/collector.yaml:/tmp/collector.yaml:Z pavolloffay/otelcol-with-mcp:0.1 --config
claude mcp add --transport=http otelcol http://localhost:9999/mcp --scope user

This MCP server is implemented as collector extension and connector which provides live view on the data.

There is a tool to get the collector config, schema for each component which is built into the collector and performs validation.

The returned schema is incomplete, and it does not contain fields explanation and type.

● OTLP Receiver Schema

  Component Type: otlpKind: receiverConfig Type: *otlpreceiver.Config

  Schema Structure

  {
    "protocols": {
      "grpc": null,
      "http": null
    }
  }

  Current Configuration (with defaults)

  {
    "protocols": {
      "grpc": {
        "endpoint": "0.0.0.0:4317",
        "keepalive": {
          "enforcement_policy": {},
          "server_parameters": {}
        },
        "read_buffer_size": 524288,
        "transport": "tcp"
      },
      "http": {
        "cors": null,
        "endpoint": "0.0.0.0:4318",
        "idle_timeout": 0,
        "keep_alives_enabled": true,
        "logs_url_path": "/v1/logs",
        "metrics_url_path": "/v1/metrics",
        "read_header_timeout": 0,
        "tls": null,
        "traces_url_path": "/v1/traces",
        "write_timeout": 0
      }
    }
  }