Product type · Updated
Help center for AI apps
If you build an AI product, your help center is probably the most out of date part of it. Credits and limits change, a new model behaves differently, a feature you shipped Tuesday isn't documented by Friday, and users ask why their results changed. usedocs keeps the docs in step with the product and answers from them.
TL;DR
Short answer
AI products change faster than almost any other software: new features every week, new usage limits, new models with different behavior. A help center for AI apps has to keep up. usedocs proposes edits when merged pull requests change what an article describes, turns users' unanswered questions into drafted articles, and answers from the docs with sources, refusing when they're silent.
Who this is for
Founders and small teams building AI powered SaaS, from writing and coding assistants to agents and AI features inside vertical tools, usually one to twenty people shipping weekly or faster. Typical shape: usage based plans with credits or limits, settings that expose model choices, and users who need to understand what the product can and can't do. Best fit when users learn the product from docs and you ship faster than anyone can write.
Problems this solves
- You ship every week, and the help center describes features, settings, and limits from two releases ago.
- Users ask why their results changed after a model update, and there's no article explaining what changed.
- Credit costs and usage limits change with pricing experiments, and old numbers linger in articles and old answers.
- Users ask what happens to their data and whether it's used for training, and the answer is buried in a policy page.
- Your own support bot, built on a general model, answers questions about your product with confident guesses.
How usedocs works for this job
Bring in your docs Import your current docs site or help pages; articles arrive as drafts. If you're starting from nothing, usedocs can draft first articles from your repository.
Connect GitHub Each merged pull request and release is screened. When it changes a feature, a limit, a default, or a setting an article describes, you get a proposed edit with the change attached.
Answer from docs, not from the model's memory The assistant answers only from your published articles, links the source, and says so when the docs don't cover a question. That matters more for an AI company than anyone.
Draft what users ask for Unanswered questions and “not helpful” votes become ranked gaps with drafted articles, so new behavior gets documented as soon as users notice it.
What you get
Docs that move at release speed
Proposed edits from merged pull requests, plus a scheduled check of every live article against the code.
Limits and credits kept consistent
When a usage limit or credit cost changes in code, the articles that state it get proposed edits instead of quietly going wrong.
An assistant that doesn't improvise
Answers come from your articles with a link to the source; weak matches get an honest no and become gaps.
Readable by other AI tools
Every page has a Markdown copy, the help center publishes llms.txt, and a Copy page menu opens articles in ChatGPT, Claude, or Perplexity, or connects them to Cursor.
Document behavior, not just buttons
AI products confuse users less through their interface than through their behavior: why a result changed, what a setting does to the output, what the product won't do. Write short articles for each, and let unanswered questions tell you which behaviors need explaining. When a release changes a default or a model, the proposed edit names the change, so the article explains it before users file a ticket.
Keep one source for limits and credits
State each usage limit and credit cost in one article and link it everywhere else, including your pricing page and in app messages. When a pull request changes a limit in code, usedocs proposes the edit to that article, and every answer that cites it is right again the moment you approve.
Why teams pick usedocs
- Built for teams that ship faster than they can document
- 7 day free trial with no card, then one plan at $99 a month
- 2,000 AI conversations a month included, follow ups free
- Nothing publishes without your approval
Not for you if
Skip this if your product is mainly an API for developers who need an interactive reference; a developer docs platform fits better, with usedocs optional for the customer help center. Also skip it if you want a general purpose agent that takes actions in your product; usedocs answers from docs and hands off.
FAQ
Why do AI apps need a different approach to docs?
They change faster: features, limits, credits, and model behavior shift weekly. Docs that are updated by hand fall behind within a release or two.
How does usedocs keep up with weekly releases?
It reads each merged pull request and release and proposes edits to the articles they affect, with the change attached, and checks every live article against the code on a schedule.
Will the assistant make things up about my product?
It answers only from your published articles and links the source. When the docs don't cover a question, it says so and hands off instead of guessing.
Can it explain changes in model behavior?
It can answer from an article that explains the change. When users ask about a change with no article, the question becomes a gap with a drafted article for you to review.
How should I document data and training policies?
In a short, plain article in the help center, linked from your policy page, so the assistant can cite it and users find it in search.
What does it cost?
One plan, Growth at $99 a month, with 2,000 AI conversations, 5 seats, and a custom domain, after a 7 day free trial with no card.