Who it is for
Small teams that ship an SDK, API or developer tool
You have public docs and a published package, but no one whose job is to make sure AI tools describe it correctly. You want the files that tools like Cursor, GitHub Copilot and Claude Code read, written from your real docs and your real package, without a project of your own to build them.
What you get
Files that point AI tools at the right answers
llms.txt and llms-full.txt
A short, organized list of your docs pages for AI tools to start from, plus a single text file with the full docs. They go at the root of your docs site.
A clean docs index
Every docs page with a one-line summary, the public functions and classes in your current release, and the things that were renamed or removed.
Agent rules files
AGENTS.md, CLAUDE.md, Cursor rules and GitHub Copilot instructions for your repository, all written from the same reviewed source, so every tool gets the same guidance.
A before and after check
We ask several AI models the same questions about your SDK, first on their own and then with the new files, and check every answer against your published package.
A plain report
What we read, which models we asked, what they got wrong, and where each file goes. No dashboard to log into.
Delivered your way
As files you drop in, or as a pull request to your repository. You own everything we deliver; nothing depends on us staying involved.
Pricing
Pick the level that fits
Starter Files
$750 one time
One SDK in one language. llms.txt, llms-full.txt, the docs index and all four rules files, plus the before and after check on a short set of questions.
Full Fix
$2,500 to $5,000
Everything in Starter Files, hand reviewed line by line, with a list of the docs pages to fix, a 30-minute walkthrough and a re-check 30 days after delivery. The price depends on the size of the SDK and the number of languages.
Monthly Re-check
$250/month
We ask the same questions every month, refresh the files when you ship a release, and send a short note on what changed. Stop any month.
Billed by Stripe invoice, due in 15 days. We agree on scope on a short call first, then send the invoice. Already a VentCheese customer? We start from your report's findings.
How it works
What happens, and when
- You send a request. Your docs link, your package name and your repository. It takes about two minutes.
- A short call. We agree on which SDK, which languages and which tier, and you get the invoice.
- We do the work. We read your public docs and published package, write the files, and run the before and after check.
- You get the files and the report. Starter Files usually arrive within 5 business days of the call, and the Full Fix within 10.
- You publish them. Or we open a pull request for you to review and merge.
What we need from you
- The link to your public docs
- Your package names, such as the npm or PyPI name
- Your public repository, or read access if you want a pull request
- Anything you already know AI tools get wrong, if you know of any
- One person who can review the drafts
We never need passwords, API keys, customer data or private code.
Questions
Common questions
Which AI tools read these files?
AGENTS.md is read by a growing list of coding agents. Cursor reads its rules folder, GitHub Copilot reads its instructions file, and Claude Code reads CLAUDE.md. llms.txt is a public proposal for a docs map that AI tools can read. We write each one in the format that tool expects.
Will this make AI tools always get our SDK right?
No one can promise that. What we can do is give the tools a clear, current source and show you, with your own before and after check, whether it helped.
Which models do you test?
A small set of current models we have access to. The report lists every model by name, and how many questions each one answered.
Our docs platform already makes an llms.txt. Is this still useful?
Often, yes. We review what it produces, and the rules files and the check are separate from it. If your existing file is already good, we will say so.
Do you need access to our private code?
No. We work from your public docs and your published package. We read the published files; we do not run them.
Do you publish what you find?
No. Your report goes only to you. We do not name customers or share their results.
How is this different from VentCheese?
VentCheese finds what AI tools get wrong about your SDK. AI Docs Fix writes the files that help them get it right. You can buy either one on its own.
Free
Free AI docs check: see what AI assistants get wrong about your docs
Send your docs link and product name. We ask a few AI coding assistants common questions about your product, check the answers against your public docs, and email you a short note on what they got wrong. No call and no invoice.
Ready to start?
Send your docs link and package name. We will reply by email to set up a short call.