Skip to main content
Glama
nixkeeper-mcp
by nixkeeper

nixkeeper-mcp

An MCP server that lets an AI assistant answer questions about nixpkgs packages from nixkeeper's data: is a package outdated, does it build on Hydra, what did the update bot make of it, which PRs and issues touch it, which CVEs affect it, and what needs attention for a maintainer or a team. It runs on your computer over stdio, downloads only public files, and needs no account or token.

Using it

Install it once, then tell your assistant to start it. With Nix:

nix profile install github:nixkeeper/nixkeeper-mcp

And with Claude Code:

claude mcp add --scope user nixkeeper -- nixkeeper-mcp

Then ask: "what's wrong with wesnoth?", "which of iedame's packages need attention?", "what broke on Darwin in the last day?", "which open PRs could the merge bot merge for the Gaming team's packages?", "which of my packages break when staging-next is merged?"

Any MCP client that starts a server on stdio works the same way, with the command nixkeeper-mcp. nix profile upgrade nixkeeper-mcp brings it up to date. Without Nix: python3 -m nixkeeper_mcp in a checkout (Python 3.11 or later, standard library only).

With Home Manager

The flake has a Home Manager module. It adds the server to Home Manager's shared MCP servers (programs.mcp.servers), which the assistants Home Manager configures read when their enableMcpIntegration is on: Claude Code, GitHub Copilot CLI and others. The server is the one your flake lock pins, so there's no claude mcp add, and nothing to download or build when an assistant starts it.

# flake.nix
inputs.nixkeeper-mcp = {
  url = "github:nixkeeper/nixkeeper-mcp";
  inputs.nixpkgs.follows = "nixpkgs";
};

# Home Manager
imports = [ inputs.nixkeeper-mcp.homeModules.default ];
programs.nixkeeper-mcp.enable = true;
programs.claude-code.enableMcpIntegration = true;

The module's options are the server's: ttl, offline, cacheDir, and dashboard, digestOwner and page for another instance's data (see Options below); name is the server's name for the assistants (nixkeeper), and package the build to run. The module is also homeManagerModules.default, for setups that look for that name.

To try it first

claude mcp add --scope user nixkeeper -- nix run github:nixkeeper/nixkeeper-mcp

This runs it without installing anything, but Nix fetches the flake on every start, and asks GitHub again once its copy is an hour old. The first start downloads and builds it, which can take longer than an assistant waits for a server to start (30 seconds in Claude Code): if it gives up, start it again once the build is done, or install it as above.

Related MCP server: SecureLLM MCP Server

The tools

Tool

Answers

From

package

Everything about one package: nixpkgs' version against upstream's, master's, Hydra's builds on each platform (why a failure failed, when it last built, platforms it's listed for but never built on), the update bot's latest attempt and when it tries next, its update PR and other PRs and issues, CVEs with nixkeeper's verdict against nixpkgs' version, maintainers and teams; an update merged into staging or staging-next on its way to master, and whether staging-next breaks the package; and what the hourly digests have that's newer than the dashboard's last sync

the dashboard's row, the builds and update-bot digests

find_packages

Packages by name, with their status

the dashboard

needs_attention

What needs attention (failing, outdated, vulnerable, a fix to backport), all of nixpkgs or one maintainer's, team's or list's, by kind and platform, worst first; or what breaks on staging-next (a forecast, its own kind); for a maintainer or team, also the PRs they can merge now with the merge bot (canMerge). Or the page's search as a query (is:builds on:darwin team:Gaming wine), with the page's link for it

the dashboard's views, digest-prs

overview

All of nixpkgs: counts and trends' latest, the page's "nixpkgs in numbers", the newest and longest-standing failures, the dependencies blocking the most builds, why builds and the bot fail, what was fixed this week

the dashboard

build_failures

Hydra builds of master that aren't ok: by package or set, system, reason, the dependency that blocked them, since when; newest or longest failing first

digest-builds

pull_requests

Open nixpkgs PRs: touching a package, by author, bucket, base branch; ready for the merge bot, those a maintainer can merge with it (can_merge), blocking the update bot, superseded

digest-prs

issues

Open nixpkgs issues about a package; build failures that build now, update requests already done

digest-prs

update_attempts

The update bot's latest attempts: one package's with its place in the queue, or by outcome, reason, since when

digest-update-bot

vulnerabilities

One package's CVEs and advisories, one CVE's packages, or the worst affecting something now

digest-vulnerabilities

data_status

When each source was last made and read, and what's cached

every meta.json

needs_attention's query takes the words of the dashboard page's search box: is: one card (outdated, failed, builds, updates, vulnerable, backport, broken, staging-next-breaks), is:unbuilt, -is:broken, has:pr, -has:pr, -has:maintainer, on:linux or on:darwin, in: a package set, team:, by: or @handle, and words of a name; without is:, every package of the maintainer's or team's, as the page lists them. A word it can't apply (age:, sort:, status:...) is refused rather than left out, and each answer links the page with the query (query.page), which takes them all.

Every answer says when its data was read (readFrom, times in UTC), and says so when a source couldn't be reached and the answer comes from what was cached. nixkeeper's checks (an issue that looks done, a PR ready for the merge bot, an attempt that looks superseded, a package staging-next will break) are hints for a person to check, not decisions: the server only reads, and never comments on, closes or merges anything. Its tools say so to the client (readOnlyHint), which may then run them without asking.

Where the data comes from

  • nixkeeper.com's data (/data/, the community dashboard: every nixpkgs package, synced every 3 hours): each package's row, with everything nixkeeper joined from the digests and its verdicts (outdated, failing, vulnerable, to backport), so a package's state reads as the page shows it. Its format: nixkeeper's docs/data.md.

  • The digests (the data branch of each repository, each README saying its format), fresher and searchable across nixpkgs: digest-builds (Hydra, hourly), digest-prs (open PRs and issues, hourly), digest-update-bot (nixpkgs-update, every 3 hours), digest-vulnerabilities (the NixOS security tracker and OSV, hourly), and digest-versions' metadata (Repology's versions are in the dashboard's rows).

A file is downloaded the first time a tool needs it, so a question about one package fetches that package's shard of the dashboard (about 90 KB) and the two digests it's compared with, not all of nixpkgs. Files are kept in $XDG_CACHE_HOME/nixkeeper-mcp (~/.cache/nixkeeper-mcp by default; about 35 MB with every source) and used for 15 minutes; after that, the server asks whether a file changed (its ETag: an unchanged file isn't downloaded again). If a source can't be reached, the cached file is used and the answer says so. Parsed files stay in memory while the server runs: up to about 400 MB with every source loaded at once (Hydra's 217,000 builds are kept as text and searched, not parsed: parsed, they alone would take that much). Every request names this repository in its User-Agent.

Options

nixkeeper-mcp [--cache-dir DIR] [--ttl MINUTES] [--offline] [--call TOOL [JSON]]
  • --cache-dir (or NIXKEEPER_MCP_CACHE): where files are kept.

  • --ttl: minutes a file is used before asking whether it changed (default 15).

  • --offline: never download; answer from the cache.

  • --call TOOL '{"name": "wesnoth"}': run one tool and print its answer, to try them without an assistant.

NIXKEEPER_MCP_DASHBOARD (default https://nixkeeper.com/data/) and NIXKEEPER_MCP_DIGEST_OWNER (default nixkeeper) point it at another instance's data, a fork's; the answers link to the page beside the dashboard's data (https://nixkeeper.com/), or NIXKEEPER_MCP_PAGE.

With Claude Code, options go after the command: claude mcp add nixkeeper -- nixkeeper-mcp --ttl 60.

The protocol

The server speaks MCP over stdio, both its current revision (2026-07-28: no handshake, each request carrying its version in _meta, and server/discover) and the earlier ones (2025-11-25 back to 2024-11-05, with initialize), whichever the client opens with. Tools only: no resources or prompts. It's written with the standard library alone (nixkeeper_mcp/protocol.py), like the digests, so it runs with any Python 3.11 or later and has nothing to install.

Development

nix flake check runs the tests (which never ask the real sources), lint and the formatting check; nix fmt formats; nix develop has the tools.

Releases: bump the version in pyproject.toml and nixkeeper_mcp/__init__.py, move CHANGELOG.md's Unreleased entries under the new version, and once that's merged, push a tag vX.Y.Z on main's commit: .github/workflows/release.yml checks it and publishes the GitHub Release with that section as its notes.

AI & automation disclosure

This server's code and documentation are written with the assistance of an AI (Claude, by Anthropic), with a person in the loop: nixkeeper's maintainer decides what to build, and reviews and tests every change before it's merged. It reads public data only, and never comments on, closes or merges anything in nixpkgs.

License

MIT

Available Tools

10 tools
build_failuresHydra build failuresA
Read-only

Search Hydra's builds of nixpkgs master (the builds digest, hourly) that aren't ok: failed (the package's own failure, with why from its log and an excerpt), dependency (blockedBy: which dependency failed), unfinished, queued. Newest first, or longest failing first. With the last success of each.

Inputs: sort, limit, since, reason, +5 more.

data_statusData freshnessB
Read-only

When each digest and the dashboard were last made and read (their meta.json files), and what this server has cached.

find_packagesFind packagesA
Read-only

Find nixpkgs packages by name (exact, then starting with, then containing the query), each with its status: failed, outdated, up to date, vulnerable.

Inputs: query (required), limit.

issuesOpen nixpkgs issuesA
Read-only

Search nixpkgs' open issues by title (the PR digest, hourly). Build-failure issues have hydra: what Hydra builds now say (builds: a candidate to close; failing; waiting; removed; ...); update requests have update: whether nixpkgs has the version asked (done, partly, open) and the update bot's plans.

Inputs: kind, limit, search, package, +1 more.

needs_attentionPackages needing attentionA
Read-only

Packages that need attention, as nixkeeper's dashboard counts them: a build failing on Hydra, outdated, vulnerable, or a CVE fix to backport; optionally only one maintainer's (GitHub handle), one team's or one list's, one kind, one platform. Without maintainer, team or list: all of nixpkgs (thousands: narrow with kind). With a maintainer or team, also the open PRs they can merge now with the merge bot (canMerge).

Inputs: kind, list, team, limit, +3 more.

overviewnixpkgs overviewA
Read-only

All of nixpkgs at a glance, from nixkeeper's dashboard: how many packages are outdated, failing to build, failing their update attempts, vulnerable, broken; the newest and longest-standing of each; the failing dependencies blocking the most builds; why builds and the update bot fail, grouped; what was fixed in the last week.

packageA package's stateB
Read-only

Everything nixkeeper knows about one nixpkgs package: its nixpkgs version against upstream's (Repology, nixkeeper's own update checks), master's version, Hydra's latest builds on each platform with why a failure failed and when it last built, the update bot's (nixpkgs-update, r-ryantm) latest attempt and when it tries next, open update PRs and other PRs and issues, CVEs and advisories with nixkeeper's verdict against nixpkgs' version, maintainers and teams; an update merged into staging or staging-next on its way to master (stagingPR, and how it builds there), and whether staging-next breaks the package (stagingNextBreaks: a forecast). newerInDigests: what the hourly digests have that's newer than the dashboard's last sync.

Inputs: name (required), repology.

pull_requestsOpen nixpkgs PRsA
Read-only

Search nixpkgs' open pull requests (the PR digest, hourly), with what the digest makes of each: its buckets (update, init, drop, tiny, bot, python, ...), the update it makes and how it compares with nixpkgs now (state), whether a maintainer can merge it with the merge bot, whether it blocks the update bot, duplicates, and packages it touches that fail on Hydra (maybe its fix).

Inputs: base, sort, limit, state, +8 more.

update_attemptsUpdate bot attemptsB
Read-only

nixpkgs-update's (r-ryantm) latest attempt at packages (the update-bot digest, every 3 hours): its outcome, versions, PR, why it failed with a log excerpt; for one package, also its place in the bot's queue and when it's expected to try again.

Inputs: limit, since, reason, search, +2 more.

vulnerabilitiesVulnerabilitiesA
Read-only

CVEs from the NixOS security tracker (with its issue, severity, affected versions, status on master and the release branches) and OSV advisories for language packages (the vulnerabilities digest, hourly): one package's, one CVE's (or NIXPKGS-..., or an advisory id), or else the CVEs affecting some package now, worst first.

Inputs: cve, limit, package, severity.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A3.6/5.0

    Scored across 10 tools

    Disambiguation4/5

    Most tools target clearly different resources or actions: find_packages searches, package retrieves one package, issues and pull_requests cover distinct artifact types, and vulnerabilities and update_attempts are specialized. There is some conceptual overlap among overview, needs_attention, and build_failures because all surface problematic packages/builds, but the detailed descriptions mostly make the boundaries usable.

    Naming Consistency3/5

    All names use snake_case, but the convention is mixed: find_packages is verb_noun while package, overview, issues, vulnerabilities, pull_requests, build_failures, needs_attention, data_status, and update_attempts are noun/resource phrases. Singular/plural usage also varies (package vs issues/pull_requests), so the pattern is readable but not strongly predictable.

    Tool Count5/5

    Ten tools is well within a reasonable range and fits the server's monitoring scope. Each tool covers a distinct area of nixpkgs health: package lookup/detail, build failures, dashboard summaries, issues, PRs, vulnerabilities, update-bot attempts, and data freshness.

    Completeness4/5

    The surface covers the main read-only monitoring lifecycle: package discovery and detail, aggregate health, attention lists, build failures, issues, PRs, CVEs/advisories, update attempts, and digest freshness. Minor gaps exist around deeper direct issue/PR detail or explicit maintainer/team listing, but core workflows appear covered.

    Maintenance

    ActivityNo data
    ResponsivenessOutdated data

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      B
      maintenance
      Access the Nix ecosystem (Nixpkgs, NixOS, Home Manager, nix-darwin, nixvim, nixhub...)
      41 PyPI
      7
      GPL 3.0
    • A
      license
      Not graded
      quality
      B
      maintenance
      Exposes code-review and issue-triage tools to AI assistants over stdio, letting them evaluate unified pull-request diffs for maintainer remarks, slice diffs into categorized AST hunks, and classify GitHub issues by priority while synthesizing minimal reproduction stubs.
      1
      Apache 2.0
    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables an AI assistant to retrieve IT system status and search or rebuild a vector-indexed knowledge base of operational runbooks. It supports local, tool-grounded responses over stdio for Ollama-based agents, helping prevent hallucinated IT operations guidance.
      MIT