Back to directory

nix-ai

A Nix flake and home-manager module that declares an entire AI coding toolkit, Claude Code, Gemini, GitHub Copilot, OpenAI Codex, Cursor CLI, Qwen Code and a shared catalog of MCP servers, as reproducible configuration, rebuilding it identically on any supported machine.

nix-ai

nix-ai captures an AI coding setup, plugins, hooks, agents, commands, rules, permissions and MCP server catalogs across several tools, as Nix code instead of manual per-machine configuration, so one command reproduces the exact same setup on a new machine. It manages Claude Code's plugin ecosystem, Gemini and Antigravity CLI and IDE settings, GitHub Copilot configuration, OpenAI Codex settings and approval policy, a Cursor CLI terminal agent, Qwen Code and Cecli settings, a shared MCP server catalog such as GitHub, Terraform, Context7, filesystem and memory fanned out to every agent, other AI dev tools such as cclint and claude-flow, and, on macOS, local Apple Silicon inference through vllm-mlx.

Added as a flake input and a home-manager shared module, it evaluates with no extra configuration required, though local MLX inference stays idle until pointed at a model. Separate exported modules let a user pull in just one tool, such as Claude Code alone, just Codex, just the MCP catalog, or just Maestro orchestration, plus a pure settings-JSON export for CI validation and a role-name-to-model-ID registry for other tooling to reuse.

Built with Nix, requiring Nix with flakes and home-manager on aarch64-darwin or x86_64-linux. MIT licensed.

More in Agent skills

See all 581
  • Provenance gate

    github.com

    A deterministic authorization gateway that sits between an AI agent and its tools, tracing where each parameter of a side-effect call came from, and blocking the call or asking a human when the source does not check out.

  • A markdown framework of system rules that restates Isaac Asimov's Four Laws of Robotics as a set of prioritized loss calculations for large-language-model agents to follow, published as a discovery file for agents and crawlers to read.

  • Vitruvius

    github.com

    An engineering research agent that runs a discover, read, synthesize, verify and review loop over technical questions, reading sources directly rather than inferring from titles, and publishing its own scored verifier benchmark, weaknesses included.