cyberneticlibrary

Audit LLM traces and experiment metrics

langsmith-mcp-servermcp_serversetup L20
npm:langsmith-mcp-server
What it does

Fetch LangSmith traces, prompts, datasets and experiment metrics

Best for

Use when you need to audit LLM traces, pull prompt templates, or compare experiment performance.

Inputs
  • · Project names
  • · Dataset IDs
  • · Thread IDs
  • · Run UUIDs
Outputs
  • · Run traces with metadata
  • · Prompt templates
  • · Dataset examples
  • · Experiment metrics
Requires
  • · LANGSMITH_API_KEY
  • · Optional: LANGSMITH_WORKSPACE_ID
Preconditions
  • · LangSmith account
  • · Node.js 18+
Failure modes
  • · Character-budget pagination requires manual page iteration
  • · Large traces may truncate strings
Trust signals
  • · MIT license
  • · Full parity with official Python implementation
  • · npm published