Explain My BuildTurn the app you built into a guide your users can follow.
You shipped something real with AI. Explain My Build reads your actual code — no screen-recording — and writes a plain, task-by-task guide the people who use your app can follow. You also get a private runbook to operate it yourself and a fix-list you can hand straight to your AI. Every line is backed by your real code, not a guess — and connect your CI and it also grades your build and flags what got worse since your last scan.
Paste a public GitHub link and get your guides in about a minute — no sign-up. Private repo or sensitive code? Run it entirely in your browser (or the CLI) with your own AI — your code never leaves your machine.
Paste a link. One button. Get a draft to review.
No setup, no jargon, no command line. Paste your app's GitHub link — or, for private code, run it right in your browser.
Paste your app's link
Paste a public GitHub link. (Private repo or sensitive code? Choose a local folder in your browser instead — nothing leaves your machine.) In about two seconds it tells you what your app is — free.
Press one button
Press "Explain my build." It reads your code and your build history and tells you what it's doing in plain words the whole way — no black box.
Get your guides
Out come a users' guide for the people who use your app, a private runbook for you, and a punch-list of fixes you can hand straight to your AI.
Review, then share
Review the draft against your app — it flags what it couldn't verify, so it's a quick pass, not a rewrite. Then publish the users' guide as a link or an in-app Help widget (Pro), or download any of them as a file.
Two guides and a fix-list from one look at your app.
🔧 A private founder runbook
How to operate, extend, and safely change the app you built — including the parts the AI wrote that you don't fully understand yet. Written for you, in plain words, so you can change one thing without breaking another. Free.
📘 A public users' guide
You review and approve it first — it's a strong draft that flags what it couldn't verify, not a finished document. Then hand it over as a link — or drop in a Help button so it opens right inside your app — and it stays current as your app changes.
🛠️ A fix-list for your AI
A ranked punch-list of verified improvements — each one checked against your real code so false alarms are filtered out — with why it matters and a copy-paste prompt that tells your AI to research the current fix and make it. Plus the setup jobs only you can do. Free.
It cites. It doesn't guess.
Most tools that "explain your AI code" just guess and hope. This one is built to be honest about what it knows.
It never makes things up
Every claim in your guides points back to something real in your app — a specific file, route, or line of code. If it can't back a claim up, it leaves it out instead of inventing it.
It tells you how sure it is
It rates its confidence in plain words and hands you an explicit "here's what I couldn't verify" checklist — so you know exactly what to double-check before you publish, and your review is minutes, not hours.
Careful with your code
Paste a public repo and we read it on our server and then delete it — we don't store your code, and only a small, de-identified summary goes to the AI. Got private or sensitive code? Run it entirely in your browser (or a local model) and nothing leaves your machine at all. Careful by design, even for sensitive apps.
Run it in your CI — it grades your build and catches backsliding.
The guides are the wedge; a real audit makes them trustworthy. Run Explain My Build in your own CI (or hand it an Auger / dh-audit report) and it grades your build, keeps your scan history, and flags anything that got worse since your last scan — all while your code never leaves your infrastructure.
A real grade, from a real audit
A ship-verdict, an overall score, and findings grouped by dimension — read straight from the audit report you provide, with skeptic-refuted false positives already removed. A clean grade means nothing matched the checks that ran; it is not a safety guarantee, and anything that couldn't be analyzed is listed as unknown, not cleared.
What got worse since last scan
Run it on every push and it tracks findings across runs — new, resolved, and regressed: a previously-fixed issue that came back because a re-prompt overwrote the fix. That's the signal a one-shot local scan can't give you.
Your code never leaves your infra
The CI path runs entirely inside your own pipeline. Only a small, de-identified scan metadata summary is posted — never your source — and the ingest token can only write scans for one repo. How we handle your code →
Connect your CI in one step.
Add the audit step to your workflow, drop in a repo-scoped ingest token, and your history + regression alarm start filling in. Zero standing access.
# In your CI, after your audit writes report.json — the app gives you the exact snippet + token:
- name: Post scan metadata to Explain My Build # metadata only — never your source
env:
EMB_INGEST_TOKEN: ${{ secrets.EMB_INGEST_TOKEN }} # repo-scoped, least-privilege
run: node scripts/emb-ingest.mjs report.json
See a real guide it made.
A genuine users' guide, generated from a real demo app's code — nothing hand-written.
For founders who built something real with AI.
For non-technical and semi-technical founders who built a real app with AI and either don't fully understand it, or need to teach their customers how to use it. It is not a "learn to code" tool. If you can pick a folder and press a button, you can run it.
Start now — it's live.
Paste a public GitHub link — no sign-up, no API key — and get your first guides in about a minute. Private or sensitive code? Run it entirely in your browser (or the CLI) with your own AI, so nothing leaves your machine. The founder runbook and fix-list are free.
🌐 Works in your browser
Paste a link — or, in Chrome, pick a local folder — and press one button. Built for people who don't live in a terminal. Nothing to install.
⌨️ Or a local CLI
Prefer the terminal? The same engine runs as a local command — handy for a larger app, or for pointing it at a fully local model.
Start free. Pay only to teach your users.
Founder runbook + fix-list
- The private runbook for your own app
- How to operate, extend, and safely change it
- A ranked, verified punch-list of fixes for your AI
- Private & local options — run sensitive code entirely in your browser
Users' guide & sharing
- The public guide that teaches your users
- An in-app Help widget — users open it right inside your app
- Share by link with an access code
- Export a portable file — and it stays auto-current as your app changes
Start free with the runbook + fix-list. Add the users' guide, sharing, and export for $19/month whenever you're ready — cancel anytime. Continuous audits and regression tracking run in your own CI and enrich the same free runbook and fix-list — no separate tier.
Not sure about a subscription? Get a one-time Guide Pass — $29 for 7 days of full access to the users' guide, no renewal. Get the Guide Pass →
Good questions.
Is my code uploaded anywhere?
It depends which path you pick. Paste a public GitHub link and we fetch it on our server to read it, then delete it — we don't store your code, and only a small, de-identified summary goes to the AI. Prefer to keep everything local? Choose a folder or use a private repo and it runs entirely in your browser with the AI you pick (Claude, OpenAI, Gemini — or a local model that sends nothing at all) — your code never leaves your machine. Either way it's careful by design, even for health-data apps — it isn't a formal BAA, so give the guides a quick read before you publish them.
What apps does it work on?
Apps you built with AI tools like Lovable, Cursor, v0, Bolt, Replit, or Claude. You just point it at the folder your app lives in.
Which AI does it use — do I need Claude?
The instant path (paste a public link) uses ours, so you need nothing. Prefer your own? Open Advanced and bring a Claude, OpenAI, or Gemini key — or a local model on your own machine, where your code never leaves it.
Do I need to know how to code?
No. If you can pick a folder and press one button, you can run it. It explains each step in plain words as it goes.
How is this different from other "explain my AI code" tools?
Most of them guess. This one cites. Every line points back to real evidence in your app, it tells you how sure it is, and it leaves out anything it can't back up.
What do I actually get?
Two plain guides plus a fix-list, from one look at your app: a private runbook for running and safely changing it yourself, a public users' guide for your customers, and a ranked punch-list of improvements you can hand straight to your AI.
Does it run a security audit or track regressions?
The instant paste-a-link path gives you the guides and a fix-list from your code and its history. The audit grade and the "what got worse since last scan" signal come from the CI-native path — run Explain My Build in your own CI (or feed it an Auger / dh-audit report), where your code never leaves your infrastructure and only de-identified scan metadata is posted. See Continuous audits.
How much does it cost?
The founder runbook and fix-list are free. The users' guide, sharing by link + access code, and portable export are $19/month — or $12/month for the first 100 founders. Billed monthly, cancel anytime.