traceloop/opentelemetry-mcp-server

Overview

This is a Model Context Protocol server that provides a unified interface for querying OpenTelemetry traces from multiple backends such as Jaeger, Tempo, and Traceloop. It enables AI agents to programmatically access distributed trace data for automated debugging and observability analysis.

Best for

Best for
Developers building AI agents for observability and distributed tracing

Use cases

• Automated root cause analysis from trace data
• AI-driven debugging of distributed system failures
• Unified trace querying across different observability backends

How to use

Install

uvx opentelemetry-mcp --backend jaeger --url http://localhost:16686

Tools exposed

• search_traces
• search_spans
• get_trace
• get_llm_usage
• list_services
• find_errors
• list_llm_models
• get_llm_model_stats
• get_llm_expensive_traces
• get_llm_slow_traces
• BACKEND_TYPE
• BACKEND_URL
• BACKEND_API_KEY
• BACKEND_TIMEOUT
• LOG_LEVEL
• MAX_TRACES_PER_QUERY

Tested with

Claude Desktop, Claude Code, Cursor, Windsurf, ChatGPT

Example client config

{\n  "mcpServers": {\n    "opentelemetry-mcp": {\n      "command": "pipx",\n      "args": ["run", "opentelemetry-mcp"],\n      "env": {\n        "BACKEND_TYPE": "jaeger",\n        "BACKEND_URL": "http://localhost:16686"\n      }\n    }\n  }\n}

Notes

This is a Model Context Protocol server that provides a unified interface for querying OpenTelemetry traces from multiple backends such as Jaeger, Tempo, and Traceloop. It enables AI agents to programmatically access distributed trace data for automated debugging and observability analysis.

188 stars on GitHub. Last updated 2026-04-20. Licensed Apache-2.0.

Use cases

• Automated root cause analysis from trace data
• AI-driven debugging of distributed system failures
• Unified trace querying across different observability backends

Pros

• Connects to multiple OpenTelemetry backends through a single MCP interface
• Designed specifically for AI agents to integrate with trace data
• Open source with a Python codebase for easy customization

Cons

• Relies on existing OpenTelemetry instrumentation in the target systems
• Requires an MCP compatible AI agent to leverage the server
• Relatively small community with 188 stars on GitHub

Indexed from awesome-mcp-servers-punkpeye and enriched against its public facts.