get_llm_expensive_traces
Find highest token usage
How to use it
get_llm_expensive_traces is exposed by the Opentelemetry MCP Server MCP server. Add the server to your MCP client (Claude Desktop, Cursor, Windsurf and others), and the get_llm_expensive_traces tool becomes available to the model automatically. See the full listing for setup details and every tool this server provides.
Install Opentelemetry MCP Server
uvx opentelemetry-mcp --backend jaeger --url http://localhost:16686Other tools in Opentelemetry MCP Server (18)
Find traces with errors
Model name (gpt-4, claude-3-opus, etc.)
LLM provider (openai, anthropic, etc.)
Get model performance stats
Find slowest operations
Aggregate token usage metrics
Get complete trace details
Filter by error status
Max results (1-1000, default: 100)
Discover models in use
List available services
/ maxdurationms - Duration filters
Filter by operation
Search individual spans
Search traces with advanced filters
Filter by service
/ endtime - ISO 8601 timestamps
Custom tag filters