ShipDocs is live: AI-generated, chat-ready docs for any codebase

Today we're launching ShipDocs - an AI-powered documentation tool that scans any codebase and generates a complete, structured, chat-ready documentation set in seconds. No template wizards. No empty Markdown skeletons. No "go write your README." You connect a repo, and you get docs.

If you've ever come back to a side project after six months and tried to remember what worker-2.ts actually does, this is for you. If you've ever onboarded a new engineer onto a 200k-line monorepo with a one-paragraph README, this is for you. If you've ever pasted your repo's tree into ChatGPT and asked "what does this even do," this is especially for you.

The README problem

Almost every codebase older than 18 months has the same problem: the docs are out of date, partial, or both. The reasons are not mysterious.

• Writing docs is slow. Writing correct docs is slower.
• Nobody on the team has "documentation" in their job title.
• The fastest way to get a missing answer is still to ask the engineer who wrote the code, which means the senior people on a team get interrupted constantly.
• LLMs make this worse, not better, because pointing Claude or GPT at a raw repo and asking "what does this do?" produces confidently wrong answers if there's no structural signal in the code.

Existing docs platforms focus on publishing - they assume you've already done the writing. They don't solve the prior question: who is going to write the docs in the first place? That's the gap ShipDocs closes. We generate the docs from your repo automatically, host them, and ground AI chat in them - and the v2 / v3 roadmap covers everything else (versioning, public docs portal, multi-language, team seats, SSO, dependency diagram, Slack/Teams).

ShipDocs is built to be the one tool a team needs for codebase docs.

What ShipDocs does

You connect a repo (GitHub, GitLab, Bitbucket [coming soon]), upload a zip, or push your local working directory from the CLI:

curl -fsSL shipdocs.sh | sh

ShipDocs scans every file, detects the real components from the file tree and the manifests it finds, and produces three things:

1. An Overview document. Tech stack, architecture, entry points, the key concepts a new developer needs first. This is the doc you hand a new hire - or paste at the top of a coding-agent context window.
2. One Markdown doc per detected component. Backend, frontend, workers, CLI, migrations, SDK, webhooks, infra - whatever your repo actually contains. Each doc lists the component's exports, its dependencies, and how it fits into the whole.
3. An AI chat grounded in those docs. Every answer cites a file path. If the answer isn't in the docs, the chat says so. No invented function signatures, no hallucinated APIs.

You can edit any of the generated docs in a WYSIWYG Markdown editor and the source .md stays clean. You can flip a project to public and get a permanent showcase URL anyone can browse - see shipdocs.sh/showcase for live examples on real OSS projects.

What's in v1

Here's what's available the moment you sign up today:

• One-command CLI that uploads your current working directory in seconds.
• Multi-source import - GitHub, GitLab, Bitbucket (coming soon), zip upload, or CLI. Public and private repos.
• Auto tech-stack detection - 250+ languages and frameworks, with local run instructions generated automatically.
• AI-generated docs - Overview plus per-component breakdown.
• Codebase chat - answers cite exact file paths, refuse to invent APIs.
• WYSIWYG Markdown editor - edit generated docs in a rich visual editor with live Markdown output.
• Public showcase pages - opt-in, permanent URLs for any project. Browse other public runs at /showcase.
• Secure auth - magic-link email, OAuth (GitHub, GitLab, Bitbucket [coming soon]), 2FA on every sign-in, multi-device CLI tokens.
• AES-256-GCM at rest with per-project keys wrapped by a master key in a KMS-backed vault.
• Your code is never used to train AI models. Not by us, not by the model providers we use.
• Markdown-first surface for AI agents - every page available as text/markdown via content negotiation, plus an llms.txt index at the root.

Pricing

The free tier is real: two repos, full structured docs, no time limit, no credit card. It's enough to feel the difference on the projects you actually maintain.

What's next

We're shipping fast. The v2 roadmap is on the public landing page and includes:

• Team members & permissions for org accounts.
• Public docs portal with templates so you can publish ShipDocs runs as full documentation sites with your own branding.
• Doc versioning tied to git refs.
• Scheduled and on-push auto-regeneration so docs stay current automatically.
• Multi-language docs via DeepL.

Beyond v2, we're working on a visual dependency diagram, Slack and Microsoft Teams integration, and SSO/SAML for enterprise.

Why we built it

Two reasons.

The first is selfish: every time we picked up an old codebase or onboarded onto someone else's project, we paid the same tax - hours or days of reading source to build a mental model that should have been a 10-minute read. We wanted that tax gone.

The second is structural: AI-assisted development is reshaping how engineers work, and the bottleneck has shifted. The model is not the bottleneck anymore. Context is. An AI assistant with a complete, structured map of your codebase outperforms an AI assistant with raw source by a wide margin. ShipDocs builds the map.

We want every engineer, on every codebase, to have that map by default.

Try it

Start a free trial on the homepage — no credit card required.

• Get started → shipdocs.sh
• CLI: curl -fsSL shipdocs.sh | sh
• See real output before signing up: /showcase
• Free tools, no signup: /tools - README generator, complexity score, explain-in-a-tweet.

FAQ

Is the free tier really free? Yes. Two repos, full structured docs, 100 chat messages per month, no credit card, no expiration.

Do you train AI models on my code? No. Our robots.txt advertises ai-train=no to every known crawler, and ShipDocs itself does not use customer projects to train any model.

What about private repos? Supported on every paid tier. AES-256-GCM at rest with per-project keys, and your code is never used to train AI models.

Can I edit the generated docs? Yes - in a WYSIWYG Markdown editor. Your edits become the new source. Regenerating doesn't overwrite manual edits unless you ask it to.

Is this open source? Not yet. The product is hosted SaaS. We may open-source the indexer once it stabilizes.

How do I start? Go to the homepage and start a free trial. No credit card required.

Ready to try it? Get started →

• Alex