CodeVF MCP

official

CodeVF MCP lets AI hand off problems to real engineers instantly, so your workflows don’t stall when models hit their limits.

GitHub
1

What can you do with Code VF MCP?

  • Set up the MCP integration — Run npx codevf setup to authenticate via OAuth and auto-configure your AI coding client.
  • Start the MCP server over stdio — Launch codevf mcp stdio for command-based clients that connect via standard input/output.
  • Start the MCP server over HTTP — Run codevf mcp http --port 3333 to expose the server via HTTP/SSE for remote or web-based clients.
  • Override the backend URL — Pass --base-url during setup or set CODEVF_API_URL to point the CLI at a custom backend.

Documentation

CodeVF CLI

Connect Claude Code, Codex, or Gemini to live human engineers.

Install

npx codevf

Quick Start (Claude Code)

  1. npx codevf setup
  2. In Claude Code, type claude
  3. It works out of the box

Use /cvf in Claude Code to force CodeVF.

Quick Start (Codex)

  1. npx codevf setup
  2. In Codex, type codex
  3. It works out of the box

Use /mcp in Codex to confirm CodeVF is connected.

Quick Start (Gemini)

  1. npx codevf setup
  2. In Gemini, restart the client
  3. It works out of the box

Commands

npx codevf setup

Configure MCP server integration with Claude Code, Codex, and Gemini:

  • Authenticate with CodeVF using OAuth
  • Auto-configure Claude Code's ~/.claude.json file
  • Create Claude Code slash commands for CodeVF
  • Select default project for context

Override the backend URL during setup:

npx codevf setup --base-url http://localhost:3000

You can also set CODEVF_API_URL for the same effect.

codevf mcp stdio

Start the MCP server for non-Claude clients:

  • codevf mcp stdio - Launch MCP over stdio (command/args-based clients)

codevf mcp http --port 3333

Start the MCP server over HTTP/SSE for non-Claude clients:

  • codevf mcp http --port 3333 - Launch MCP over HTTP/SSE (shows endpoints)

Requirements

Support

License

Commercial license — see LICENSE.