npm package discovery and stats viewer.

Discover Tips

  • General search

    [free text search, go nuts!]

  • Package details

    pkg:[package-name]

  • User packages

    @[username]

Sponsor

Optimize Toolset

I’ve always been into building performant and accessible sites, but lately I’ve been taking it extremely seriously. So much so that I’ve been building a tool to help me optimize and monitor the sites that I build to make sure that I’m making an attempt to offer the best experience to those who visit them. If you’re into performant, accessible and SEO friendly sites, you might like it too! You can check it out at Optimize Toolset.

About

Hi, 👋, I’m Ryan Hefner  and I built this site for me, and you! The goal of this site was to provide an easy way for me to check the stats on my npm packages, both for prioritizing issues and updates, and to give me a little kick in the pants to keep up on stuff.

As I was building it, I realized that I was actually using the tool to build the tool, and figured I might as well put this out there and hopefully others will find it to be a fast and useful way to search and browse npm packages as I have.

If you’re interested in other things I’m working on, follow me on Twitter or check out the open source projects I’ve been publishing on GitHub.

I am also working on a Twitter bot for this site to tweet the most popular, newest, random packages from npm. Please follow that account now and it will start sending out packages soon–ish.

Open Software & Tools

This site wouldn’t be possible without the immense generosity and tireless efforts from the people who make contributions to the world and share their work via open source initiatives. Thank you 🙏

© 2026 – Pkg Stats / Ryan Hefner

infra-kit

v0.45.0

Published

infra-kit

Readme

infra-kit

A monorepo DX CLI (v0.3.10) that doubles as an MCP server for AI coding agents. It orchestrates local development, releases, environment management, and repo audits—both as a command-line tool and as a set of structured tools exposed to Claude and other AI agents.

Install in a Repo

Pin it as a devDependency (not global):

npm install --save-dev infra-kit
# or with pnpm
pnpm add -D infra-kit

Global installs are non-reproducible across machines and CI—always declare it in package.json.

Expose to AI Agents

If your repo uses Claude Code or another MCP-aware agent, add .mcp.json at the repo root:

{
  "mcpServers": {
    "infra-kit": {
      "type": "stdio",
      "command": "npx",
      "args": ["--no-install", "infra-kit", "mcp"]
    }
  }
}

The agent will then have access to a curated subset of infra-kit commands as structured tools (see MCP Exposed Tools below).

Other stdio MCP servers whose secrets come from ik env-load (Grafana, for example) are declared once, in infra-kit.json under mcp.<name> (command, args, env, unset). That block is the source of truth. The matching .mcp.json entry — ik-mcp --name <name> --env … -- <command> … — is generated from it by infra-kit setup and infra-kit audit --root --fix, and infra-kit audit reports drift between the two. Edit infra-kit.json, regenerate, commit both; never hand-edit the generated entry. It stays in .mcp.json by design: the proxy reads argv only, so the file Claude Code loads is self-contained and needs no config reader at session start.

Configuration

The split is by level: one infra-kit.json per repo, one infra-kit.config.ts per workspace package. The repo root has no infra-kit.config.ts.

infra-kit.json (repo)
Repo-wide settings: environment providers (e.g., Doppler), worktrees, MCP proxies, and the audit block infra-kit audit --root checks the root against. Layered with personal overrides from ~/.infra-kit/infra-kit.json and ~/.infra-kit/projects/<repo>/infra-kit.json; aliasRepo, mcp, vendorSource, vendorTarget and audit are accepted only in the committed file. A vendorSource.copy entry with a module name reaches only the targets whose vendorTarget.modules lists it (no vendorTarget block: every module); dropping a module deletes its files under vendor/ and leaves a path outside it to the target, and vendor sync fails the target while one of its package.json files still depends on a package the drop would delete. A mirrored file keeps the target's own lines between # infra-kit:local-start and # infra-kit:local-end marker lines (any comment leader): vendor sync writes the source with that block in place of the source's own block (or appended), and vendor check leaves the block out of its checksums. aliasRepo (a DNS label, e.g. "trvl") replaces the main checkout's directory name in the dev host <release>.<package>.<repo>.localhost. audit.exclude lists workspace globs in the pnpm-workspace.yaml dialect (e.g. apps/special/ui/pages/*) that audit --all and audit --fix --all skip — for packages another tool generates and owns; --json names them in excluded. A leftover root infra-kit.config.ts fails audit --root; infra-kit audit --fix --root moves its rules into audit and deletes it. audit --fix --root also stamps $schema (the JSON Schema shipped with that CLI version, on unpkg) and configVersion at the top of the committed file; nothing else writes them, and a CLI refuses a configVersion newer than it reads.

infra-kit.config.ts (package)
Strict schema file that declares the package's audit rules (requiredScripts, requiredFiles), an optional type override (frontend | backend | lib | e2e | mobile) used by the agent-guidance check, and the dev.proxy / e2e blocks read by the Vite plugin, infra-kit dev and the Playwright helper.

An e2e package (declared, or inferred from apps/<app>/tests / a @playwright/test dependency) is additionally held to the shared Playwright convention, which no config key can switch off: the e2e-test* / playwright-install scripts must carry their canonical values (e2e-script:* checks), and playwright.config.ts must read E2E_SLOW_MO, pass it to launchOptions.slowMo, add it to timeout, and keep traces on local failures (e2e-config:* checks). Each failure prints the expected line.

Commands

| Group | Command | Purpose | | ----------------- | ----------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | Develop | dev | Start local dev server with portless proxy (interactive wizard, long-running) | | | dev-status | Report what infra-kit dev currently has running (read-only; finds the fragment dir by searching upward, so it works from any subdirectory) | | Release | release merge-dev | Merge origin/dev into open regular release branches in a scratch worktree, then push atomically (--dry-run, --versions). Hotfixes (PRs to main), title/base mismatches and heads with PRs to both bases are listed as skipped with a reason. --keep-conflicts hands a real conflict to a CLI-owned resolution worktree; --continue/--abort resume or drop that hand-off. /infra-kit:merge-dev is the agent entry point for the whole flow | | | release list | List available releases | | | release create | Create a new release; -r <token>[@yyyy-mm-dd][:type[:description]] — the @date is the Jira fix version’s planned release date | | | release edit | Edit a release’s Jira description and/or planned release date (--description, --release-date) | | | release deploy-all --from <ci\|local> | Deploy every service — in CI, or from this machine | | | release deploy-selected --from <ci\|local> | Deploy chosen services — in CI, or from this machine | | | release deliver | Deliver to prod (CLI-only, irreversible) | | | release remove | Tear down one release — worktree, PR, local and remote branches, then the Jira fix version last. Refuses a merged PR, a released fix version, or one with issues attached (--move-issues-to is the opt-in). Resumable: a partially-removed release re-runs cleanly. Over MCP the Jira step is never attempted (jira: 'manual') | | Worktrees | worktrees add | Create release worktrees (--versions / --all), or feature worktrees on feature/<name> (--feature <name> [--base <dev\|release>], base dev by default) | | | worktrees list | List all worktrees | | | worktrees remove | Remove a worktree (no --force; a branch git refuses is reported and the command exits non-zero; a leftover git already unregistered that holds only .omc/state, .omc/sessions or .DS_Store is swept automatically) | | | worktrees sync | Sync worktree state (same removal and failure reporting as worktrees remove) | | Environment | env-status | Show Doppler env status | | | env-list | List secrets in current env | | | env-load | Load env secrets into shell | | | env-clear | Clear loaded secrets | | | env-token-list | List redacted service tokens | | | env-token-set [env] | Add or replace an env’s Doppler service token — masked prompt, piped stdin (op read … \| ik env-token-set dev), --from-env <VAR>, or --from-file <path> for several <env>=<token> lines; validated against Doppler before writing; no env → picker incl. a new env | | Configuration | config-get | Print the fully merged infra-kit config, or one top-level key (config-get worktrees) (read-only) | | | config-set <key> <value> | Set a top-level key in this project’s per-machine override (--global for every project); JSON or bare-string value, kept only if the config still loads | | | config-unset <key> | Remove a top-level key from the per-machine override (--global likewise) | | | config path | Print config file locations | | | config edit | Edit infra-kit.json in $EDITOR | | Vendor | vendor check | Validate vendor mirrors | | | vendor config | Show vendor manifest | | Setup | infra-kit setup [--tools <ids...>] [--update [ids...]] [--skip-tools] | Set up infra-kit on this machine: shell integration, the Claude Code plugin, the .mcp.json key, then install or update the external CLIs (brew, git, aws, gh, doppler, portless). --tools / --update narrow the converge; --skip-tools writes only, reporting what is missing without installing it. Written binary-qualified here because pnpm setup and pnpm run setup both exist in every consumer and both mutate | | | audit [--fix] [--design] | Check repo against config rules, including the per-package agent-guidance CLAUDE.md check; --fix writes guidance blocks (CLI-only, not in MCP), --design also scaffolds DESIGN.md | | | doctor [--fix] | Diagnose this machine: infra-kit install, tool auth, and repo wiring; --fix repairs what it safely can (CLI-only — the MCP tool is read-only) | | | version | Show installed version |

Agent guidance blocks (CLAUDE.md)

infra-kit audit checks each workspace package for a managed guidance block in its CLAUDE.md, marked by <!-- infra-kit:package:begin --> / <!-- infra-kit:package:end --> (the root uses its own <!-- infra-kit:begin --> / <!-- infra-kit:end --> pair). Text outside the markers is never touched.

Run audit --fix to write the block for the current package, --fix --all for every package, or --fix --root for the root file. Add --design (with --fix) to scaffold a DESIGN.md skeleton for frontend/mobile packages that lack one. Both flags are CLI-only — never exposed through MCP.

Enforcement follows workspace adoption: until any package carries a well-formed block, a missing or broken CLAUDE.md only advises; once one package has adopted, every package needs one.

infra-kit dev

Starts a local dev server with a portless-powered proxy. The interactive wizard guides you through app selection when run in a TTY without flags:

infra-kit dev              # Interactive wizard
infra-kit dev --app=client # Start one app
infra-kit dev --no-watch   # Build once; don't rebuild & restart on save (watch is on by default)
infra-kit dev client       # An app plus the backends its routes can run locally

Runs long-running until you press Ctrl-C. Manages multiple backend/frontend processes, health probes, and live proxy reconfiguration.

MCP Exposed Tools

When infra-kit runs as an MCP server (via .mcp.json), these commands are available as structured agent tools:

  • dev-status
  • gh-merge-dev, gh-release-list, release-create, release-remove, release-edit, gh-release-deploy-all, gh-release-deploy-selected
  • release-create offers an argument form when releases is omitted (one release; batches are arguments-only).
  • local-deploy-all, local-deploy-selected
  • worktrees-add, worktrees-list, worktrees-remove, worktrees-sync
  • env-status, env-list, env-load, env-clear, env-token-list
  • config-get, vendor-check
  • audit, version
  • doctor (read-only), setup (gated — every call raises the host prompt and the confirm round-trip)

26 tools total.

Not exposed: dev (long-running), release deliver (irreversible prod delivery), config edit / config path (interactive/CLI-only), vendor config (CLI-only), and mcp (process-level).


Node: ≥24.x | License: Proprietary