MCP‑NixOS – Reliable Package Knowledge for AI Assistants

MCP‑NixOS solves a common pain point for developers who rely on generative assistants: the tendency of language models to hallucinate when asked about package names, configuration options, or version histories. By exposing a lightweight, asynchronous server that queries the official Nix package registry and related services (such as Home Manager and nix‑hub.io), MCP‑NixOS provides real, up‑to‑date data. This guarantees that an AI assistant can answer questions about 130 000+ NixOS packages, 22 000 configuration options, or the precise commit hash that introduced a particular Ruby version without resorting to guesswork.

The server is built on FastMCP 2.x, which brings full async support and a clean API surface. Developers can integrate it into any MCP‑compatible workflow—whether they are using Claude, OpenAI’s ChatGPT, or a custom assistant—by simply adding the server to their MCP configuration. Once running, the assistant can invoke high‑level tools such as , , or to retrieve structured JSON results. These tools abstract away the underlying HTTP requests and data parsing, allowing developers to focus on higher‑level logic rather than boilerplate code.

Key capabilities include:

• Package and option discovery: Search by name, type (package, program, option), or channel; retrieve detailed metadata such as description, dependencies, and documentation URLs.
• Ecosystem statistics: Quickly obtain counts of packages or options per channel, helping to gauge the maturity of a particular release.
• Flake integration: Explore community‑maintained flakes with and view ecosystem metrics through .
• Version history: Access the full commit‑based timeline of a package via and locate exact versions with .
• Home Manager support: Query user‑level configuration options, including category listings and detailed descriptions.

In practice, a developer can ask an AI assistant to “find the latest stable version of that works on NixOS 23.05” or “list all Home Manager options related to SSH keys.” The assistant will forward these queries to MCP‑NixOS, receive authoritative answers, and present them in a conversational format. This reduces the risk of misconfigurations caused by hallucinated package names, speeds up onboarding for new team members, and enables automated tooling that relies on accurate dependency information.

Because MCP‑NixOS runs on any operating system—Windows, macOS, or Linux—and can be launched via simple commands (, , or Docker), it fits seamlessly into diverse CI pipelines, local development environments, and cloud‑based assistant deployments. Its lightweight design, combined with a rich set of domain‑specific tools, makes it an indispensable component for any workflow that demands trustworthy NixOS data from a generative AI.