Comet Opik
FreeComet 官方集成，用自然语言查询和分析 Opik 日志、追踪、提示等 LLM 遥测数据。
FreeFree tier
About Comet Opik
Comet Opik MCP server allows LLM engineers to interact with their Opik workspace directly from AI-powered coding tools like Claude Code, Cursor, VS Code Copilot, and MCP Inspector. It enables reading traces, logging scores, saving prompt versions, and asking Ollie (Opik's AI assistant) questions all via natural language. Built for teams using Opik for LLM experiment tracking, this MCP server integrates seamlessly into the development workflow, reducing context switching and enabling powerful debugging workflows.
Key Features
Read traces from Opik workspace
Log scores to traces with reasons
Save and manage prompt versions
Ask Ollie investigative questions about experiments and traces
List Opik projects and explore workspace data
Seamless integration with Claude Code, Cursor, VS Code Copilot, MCP Inspector
Natural language interaction with LLM experiment telemetry
Pros & Cons
Pros
- Directly integrates with popular AI coding tools like Claude Code and Cursor
- Enables natural language querying of experiment data and traces
- Open source and free to use
- Reduces context switching by keeping developers in their coding environment
- Built specifically for LLM engineers already using Opik for experiment tracking
Cons
- Requires Python 3.13+ to run
- Needs Opik API key and workspace setup for cloud users
- Currently pre-release, not yet published to PyPI (as of writing)
Best For
Debugging LLM experiment regressions by querying traces via chatScoring traces during evaluation directly from an AI assistantManaging and versioning prompts for LLM applicationsExploring Opik workspace data without leaving the coding environment