pal-mcp-server

Why we included this project

Teams that live inside Claude Code, Codex, or Gemini CLI usually end up locked to whichever model that CLI was built around. PAL removes that lock by running as an MCP server those CLIs can pull from, with Gemini, OpenAI, Anthropic, Azure, Ollama, and OpenRouter available as backends, and you can mix providers within one prompt so the model doing a code review is not the same one writing the code. Context carries across models, and a CLI-to-CLI bridge spawns isolated subagent sessions for multi-pass code review, planning, or debugging without dirtying your main context window. That makes it a practical fit for a solo developer who wants a second opinion from a different model and for small teams that would rather standardize on one gateway than juggle provider SDKs. If some work has to stay off the network, the Ollama path covers local models.