🧵 3. The setup is small.
Install neo-mcp, create a NEO secret key, register it with Claude Code, and then ask the agent to use NEO for tasks like evaluating AI Agent and optimizing it, fixing a training run, benchmarking and optimizing LLM prompts, building an ML pipeline, or documenting eval results.
🧵 2. Most coding agents are great at editing code and answering questions.
But AI/ML work often needs a longer loop: run, fail, debug, evaluate, compare tradeoffs, change the plan, and run again.
That’s where NEO fits.
Neo MCP acts as the bridge between coding agents like Claude Code, Cursor, Codex, or VS Code and NEO. The coding agent sends an AI/ML task through the Model Context Protocol, and NEO takes over the execution.
The developer stays in the same editor, while NEO handles the slower engineering loop: model evals, prompt testing, RAG debugging, benchmarks, fine-tuning, output analysis.
Then it returns what matters back into the repo: code, metrics, reports, and runnable artifacts.