Model Context Protocol

Your MCP server,
handled by NEO.

Connect Claude Code, Cursor, VS Code, and other MCP clients to NEO. The editor stays in control of the conversation. NEO does the ML engineering.

Install neo-mcp Read the docs

Two commands to connect

Claude Code (recommended)

Terminal

pip install neo-mcp
claude mcp add --scope user neo \
  -e NEO_SECRET_KEY=sk-v1-your-key \
  -- python3 -m neo_mcp

Cursor / VS Code / other clients

mcp.json

{
  "mcpServers": {
    "neo": {
      "command": "python3",
      "args": ["-m", "neo_mcp"],
      "env": {
        "NEO_SECRET_KEY": "sk-v1-your-key"
      }
    }
  }
}

How it works

Your editor routes the request

Claude Code, Cursor, VS Code, or another MCP-compatible client sends the task to the local neo-mcp server.

NEO gets the real job

NEO receives the task with enough context to plan, run code, ask for feedback, and return the finished transcript.

Results land back where you work

Status, messages, files, and follow-up questions come back through the same editor instead of a separate dashboard loop.

Built for real work

Tool routing without babysitting

NEO handles task submit, status polling, feedback, and transcript retrieval so the editor can stay focused on the conversation.

Local secrets by default

Provider keys stay on your machine and are exposed to tasks as environment variables when a run actually needs them.

One MCP bridge, many editors

The same server works across Claude Code, Cursor, VS Code, Windsurf, Zed, Continue, Codex CLI, and other MCP clients.

Transparent by design

API keys stay local and are not printed in task logs.
NEO pauses for feedback instead of guessing when it needs a decision.
Every run produces a transcript you can inspect after completion.
The integration is removable: delete the MCP entry and the editor stops calling NEO.

Ready to connect?

Install neo-mcp, add your NEO Secret Key, and start routing ML tasks from your editor.

pip install neo-mcp

Your MCP server,handled by NEO.