You push
The branch is read on every commit and its status updated. Silently: no comment, ever.
The core problem
README ■API docs ■architecture diagrams ■deployment guides ■environment variables ■the internal wiki ■release notes ■onboarding guides ■
Core philosophy
every other AI doc tool asks
“Generate the documentation.”
So you get more prose to maintain, more pages nobody asked for, and still no idea which of it is already wrong.
DocParity asks
So you fix the two files a change actually broke, each with a reason, before the drift ever reaches a reader.
The loop
DRIFT DIES
IN REVIEW,
not in production. Keep scrolling, it moves sideways.
Rules decide
Your rules, path heuristics, learned co-changes. Which docs are affected is proven, not guessed.
The PR opens
One comment, rewritten in place on every push. The answer was already computed.
An LLM drafts
Only the docs the rules flagged. It answers NO_CHANGE when nothing is stale.
You merge
The docs PR folds the fix into your branch, and closes itself if you write the docs first.
The receipt
The trust model
Overstating confidence on a guess is how doc bots die. DocParity never lets a suggestion wear a finding's badge. And when it does get one wrong, /docparity ignore closes the finding and teaches the engine to weigh that rule lower in your repo.
Who it's for
Pricing
CHEAPER THAN A
CONFUSED NEW HIRE.
Usage is metered per pull request, counted once however many times you push to it. Hit the Free tier's limit and drift detection keeps running: only the drafted fixes pause, and the bot says so rather than going quiet.