loopkit

You shipped the thing. This does the boring, repetitive launch paperwork — and only that.

You just shipped a first version of a product (an MVP — a "minimum viable product", the smallest thing worth releasing). Now you have to find users, and most of that time goes to rebuilding the same checklist from memory every time: what goes in a "Show HN" post (the Hacker News launch post) again? where does the landing-page copy come from? how do I wire up a payment processor without it tangling into my code? am I actually growing?

loopkit does that mechanical busywork — and nothing else — by transforming your own repo's files the same way every time:

npx @allenwu06/loopkit landing      # a real static landing page from your README
npx @allenwu06/loopkit launch       # Show HN / forum / changelog post DRAFTS
npx @allenwu06/loopkit checklist    # the distribution + payments wiring checklist
npx @allenwu06/loopkit track        # a plaintext, local-only traction log

No LLM. No network. No telemetry. No account. Zero runtime dependencies. MIT.

What loopkit does NOT do — by design, and this is the whole point

Read this first. It is the reason this tool exists and the reason to trust it.

• It does not do taste. It will not tell you if your value prop is any good, if your name is bad, or if the thing is worth launching.
• It does not do distribution judgment. It will not pick your channel, your timing, your audience, or your community. Those decisions are the hard part and no tool can do them for you. loopkit just stops you forgetting the steps.
• It does not post anything, ever. It writes DRAFT files to your disk. You read them, cut what isn't true yet, add your voice, and you hit submit. There is no posting code in this repo and there never will be.
• It does not write a single claim of its own. Every word in every generated artifact is either copied from your own README/CHANGELOG or is a loud [loopkit: FILL THIS IN — …] placeholder. If your repo doesn't substantiate a claim, loopkit will not invent one. It would rather emit an ugly placeholder than a plausible lie.
• It contains zero payment code. loopkit checklist generates a payments integration scaffold — documentation, the wiring steps, and TODO(you) markers. It imports no payment SDK, handles no key, processes no money. You wire that yourself, with your processor's own SDK, behind your own keys.
• It is not growth magic. There is no AI, no autopilot, no "we'll get you users". A tool that claimed that would be lying, and you should not star it.

loopkit is the boilerplate-removal of the distribution loop, not the distribution. The 80% that is mechanical (scaffolding, checklists, templates, restating your own words, tracking numbers you recorded) it does completely. The 20% that is taste, judgment, voice, and the actual clicking of "post" — it deliberately leaves to you, loudly. Every generated file says this about itself, on its own.

These guarantees are not just prose: they are statically asserted by the test suite across the entire shipped tree (test/no-network.test.js, test/gen-checklist.test.js). No network module is imported, no fetch exists, no payment SDK is present, no LLM is wired. The honesty is the moat.

The four commands

loopkit landing

Generates one static, dependency-free index.html from your README.md + package.json:

• Title and tagline from your package.json name/description (or your README's # title and intro).
• Install snippet lifted verbatim from the first fenced code block under an Install/Usage heading.
• "Why / Features / How it works / Limitations" sections, each the first paragraph of the matching README section, verbatim (markdown stripped to text — never paraphrased).
• A GitHub button only if package.json has a repository.

The file has no JavaScript, no webfont request, no CDN, no tracker, no build step. scp it to any static host and it works. loopkit does not deploy it — deployment is yours.

npx @allenwu06/loopkit landing                     # → ./loopkit-out/index.html
npx @allenwu06/loopkit landing --repo . --stdout   # print it instead

Missing a tagline or install snippet? You get a visible [loopkit: FILL THIS IN — …] in the page (you'll see it in the browser tab), never a fabricated sentence.

loopkit launch

Generates three launch-post DRAFTS from your README.md + CHANGELOG.md:

| file | what it is | |------|------------| | show-hn.draft.md | A Show HN title (with a length check against HN's 80-char limit) + a first-comment draft built from your tagline, your README feature bullets, and your own "what it does NOT do" section. The personal why you built this and your ask are left as explicit placeholders — loopkit cannot and will not write your story. | | forum-post.draft.md | A generic subreddit / Lobsters / forum post. loopkit will not choose the community — wrong-community posting is the fastest way to get flagged, and that is a judgment call. | | changelog-announcement.draft.md | A short + a long announcement built only from the top real release in your CHANGELOG.md. No changelog → it says so and emits a placeholder; it does not invent a release. |

Every draft opens with a DRAFT banner, is numbered DRAFT n/3, and restates only what your repo already says. loopkit adds no adjective you didn't write. It never posts.

npx @allenwu06/loopkit launch          # → ./loopkit-out/launch/*.draft.md

loopkit checklist

Emits the mechanical distribution + monetization checklist a solo shipper otherwise rebuilds from memory every time:

• DISTRIBUTION-CHECKLIST.md — the literal submission steps for Show HN (linking the official rules), Product Hunt, Reddit/Lobsters, owned channels, and a Merchant-of-Record monetization section. Every item is an unchecked box for a thing only you can do — loopkit checks nothing off and cannot.
• payments/INTEGRATION.md — a Merchant-of-Record (Paddle / Lemon Squeezy / Polar) integration scaffold: the wiring steps, where keys go (env vars, never committed — loopkit will not hold a key), the webhook shape and its failure modes, and TODO(you) where your code goes. It contains no runnable payment code, no SDK import, no endpoint. A generic "just works" payment handler would be a security liability — signature verification must be your code, written on purpose.

npx @allenwu06/loopkit checklist       # → ./loopkit-out/{DISTRIBUTION-CHECKLIST.md,payments/INTEGRATION.md}

loopkit track

A plaintext, local-only, git-friendly traction log. You record the numbers you observed; loopkit computes deltas and a summary. That is the entire feature.

npx @allenwu06/loopkit track add stars 12 "morning after Show HN"
npx @allenwu06/loopkit track add stars 47 "HN front page"
npx @allenwu06/loopkit track
# stars  47  [+35 total since 2026-03-01]  last step +35 (since …)

The store is a tab-separated file you keep in your own git — readable, diff-able, no lock-in. loopkit never fetches a number for you (it cannot — there is no network client). A tracker that auto-scraped your star count would be both the "magic" this tool refuses and a telemetry surface. This stays a notebook, on purpose. No phone-home, no account, no analytics.

Install / run

Requires Node ≥ 20. Zero runtime dependencies.

npx @allenwu06/loopkit help
# or, from a clone:
node bin/loopkit.js help

Common flags: --repo DIR (default: cwd), --out DIR, --stdout, --version.

How it stays honest (the design)

Everything downstream is a pure function of one normalized model (src/repo-read.js), which is itself a pure function of your package.json / README.md / CHANGELOG.md. There is no clock or RNG inside a generator, so the same repo produces byte-identical output every time (asserted by the determinism tests). The only nondeterministic thing in the whole tool is the timestamp loopkit track records — a local clock read, isolated to the CLI boundary, never a network call.

When an input is genuinely absent, loopkit emits a deliberately ugly [loopkit: FILL THIS IN — …] placeholder rather than a fabrication. That placeholder is impossible to mistake for real copy and impossible to ship by accident — which is the point.

Honest limitations

• It reads package.json, README.md, and CHANGELOG.md. It does not parse pyproject.toml, Cargo.toml, go.mod, etc. yet — if your repo is not Node-ish you get the README-driven half (still useful) and placeholders for the package metadata. (PRs / issues welcome — see below.)
• The landing page is intentionally plain (system fonts, no JS). It is a correct, deployable scaffold, not a designed site. Design is taste; loopkit does not do taste.
• It restates your README/CHANGELOG. If those are thin, the drafts are thin — honestly thin, with placeholders, which is the correct signal that the input needs work before you launch. loopkit will not paper over a thin README with invented enthusiasm.
• The checklist encodes today's platform conventions (Show HN, PH, etc.). Those platforms change; the checklist is a strong default, not a substitute for reading each platform's current rules (which it tells you to do).
• It is single-purpose on purpose. It will not become a CRM, an email tool, an analytics dashboard, or an "AI growth assistant". Those either need taste/judgment or a network/account, and both are out of scope by design.

Feedback

Misfires and "this scaffold was wrong/misleading" reports are the single most valuable input — see FEEDBACK.md. The zero-friction path: open an issue and add the loopkit-feedback label. Whatever you write is read exactly as written — not summarized, not paraphrased.

License

MIT — see LICENSE.