How do AI-generated doc PRs get into DraftView?

Connect the GitHub App to your organization once. After that, every documentation PR in a connected repo (public or private) lands automatically in a shared review queue with its own lifecycle state, so nothing gets pasted in by hand and nothing falls through. You can still load any individual public PR by URL when you just want to try it.

Do my reviewers need a GitHub account?

No. Reviewers join your organization by email with a magic link and start reviewing right away, with no GitHub account or password. A GitHub-linked teammate then pushes their feedback upstream under a real account, so every suggestion stays properly attributed. For one-off outside reviewers like legal counsel or localization, you can also share a password-protected external link.

How does DraftView know which PRs were written by AI?

DraftView reads signals from the PR itself: the author, branch name, description, and commit trailers. Detection is generator-agnostic, so it works whether the docs came from Claude Code, Copilot, Cursor, or an in-house pipeline, and it is held to a strict precision bar before a label ever drives a filter or a metric. The label flags a PR for human review. It never blocks or merges anything on its own.

What happens after a reviewer requests changes?

Their comments and edits sync to GitHub as native review threads and Suggested Changes, so GitHub stays the single source of truth. When the agent pushes a new commit, the addressed view shows the before and after of the exact region they commented on, so the reviewer can confirm the feedback was really handled, not just that a line moved.

Does DraftView replace GitHub, or lock us in?

Neither. GitHub still owns your content, your comments, and the merge button. DraftView sits on top as a review layer you can add or remove at any time, and it never auto-merges. The one thing we keep that GitHub does not produce is the human-oversight record.

Can DraftView prove a human reviewed the AI's work?

Yes. Every sign-off writes an append-only record of who reviewed, what changed, how long they spent, and against which version. You can export a per-PR oversight audit, and the dashboard reports the share of AI doc PRs that received a substantive human review versus those rubber-stamped. That is the evidence a green checkmark cannot give you, framed around EU AI Act Article 14 human oversight.

Which documentation formats are supported?

We currently support standard Markdown, MDX and AsciiDoc. AsciiDoc renders correctly, however some features may not be fully functional. We are actively working on improving AsciiDoc support. If you have specific format needs, please contact us to discuss how we can accommodate your workflow.

We use X platform for documentation. Can I still use DraftView?

We support Docusaurus, Mintlify, GitBook, Hugo, MkDocs, and any other platform that uses a GitHub-based workflow. As long as your files are in a GitHub repository, we can render the preview.

We use GitLab/Bitbucket. Can we still use your visual review engine?

Not yet. DraftView currently only supports GitHub. We are building integrations for GitLab and Bitbucket next. Contact us to request priority for your platform.

Yes, all data is encrypted at rest and in transit. We follow industry best practices for security and compliance, including SOC2 readiness. Your documentation, review sessions, and the human-oversight record are protected with enterprise-grade security measures.

Is there a free tier or trial available?

Yes. Reviewing public PRs yourself is free forever (up to 5 reviews per repo each month). For teams, connecting your GitHub organization starts a 14-day free trial of the Team plan: every doc PR lands in a shared review queue and you can invite unlimited reviewers, no GitHub accounts or credit card required. After the trial it's $29/month per organization, including 3 active reviewers.

WorkflowJune 13, 20265 min read

Your AI Opens Doc PRs Faster Than Your Team Can Review Them

AI agents open documentation pull requests at machine speed, but review still happens at human speed. The PRs pile up, age, and get rubber-stamped. DraftView organizes every AI doc PR into a review queue with an explicit lifecycle, so nothing merges without the sign-off it needs.

AI Opens the PR. The Queue Fills Up.

A year ago, a documentation pull request meant a person sat down and wrote something. Today an agent does it. GitHub Copilot, Cursor, Claude Code, Devin, and a growing list of in-house pipelines draft changelogs, regenerate API reference pages, update tutorials after a code change, and open the PR without anyone typing a word.

The drafting got faster. The reviewing did not.

Your team still has one product manager who approves user-facing copy, one legal contact for anything compliance-sensitive, and one engineer who knows whether the new API description is actually correct. They read at the same speed they always did. So the PRs arrive faster than anyone can clear them, and they collect in a queue that nobody designed.

What Breaks When Doc PRs Outpace Review

When the inflow beats the review rate, three things happen, and all of them cost you.

PRs age. A doc PR that sits for two weeks describes a product that has already moved on. By the time someone reviews it, the diff fights with three later changes.

Reviews turn into rubber stamps. When forty PRs wait and one reviewer owns them, the rational move is to skim and approve. The approval still happens. The reading does not. (More on why that is dangerous in How to Prove a Human Actually Reviewed Your AI-Generated Docs.)

Things merge without sign-off. Someone merges to unblock a release, the legal review that was supposed to happen never did, and you find out when a customer does.

None of this is a discipline problem. It is a capacity problem with no structure around it.

A Queue Built for Machine-Speed Inflow

DraftView treats the pile of AI doc PRs as what it is: a work queue that needs triage, status, and a finish line. Three pieces make that work.

Auto-ingest. Connect a repository once, and every documentation PR that opens against it shows up as a queue item. Nobody pastes a URL or remembers to add it. The PRs that touch docs land in the queue, and the PRs that do not stay out of it.

Authorship detection. DraftView reads the signals an agent leaves behind (the author login, the branch name, the PR body, and the co-author trailers in commit messages) and labels the PR with the tool that produced it. The detection stays generator-agnostic on purpose, because betting on one vendor in a market this young is a bad bet. A Dependabot version bump never gets flagged as AI. A Claude Code branch does.

A lifecycle, not a label. Every item moves through named states, so anyone can see at a glance where it stands.

The Lifecycle, Start to Finish

A doc PR in the queue moves through states that match how review actually happens:

New. The PR landed and waits for a human to pick it up.
In review. A reviewer opened it in DraftView.
Changes requested. The reviewer left comments and suggested edits.
Waiting on AI. Those comments went back to the PR as native GitHub suggestions, so the agent that wrote the docs can revise them.
Addressed. The agent pushed a commit that touches the regions you commented on. DraftView notices and surfaces it for another look.
Signed off. An approver marked the review complete.
Ready to merge. Sign-off is done and CI is green.
Merged. A human merged it on GitHub. DraftView never merges for you.

The states between Changes requested and Addressed are the part most workflows miss. AI review runs as a loop. The reviewer asks for a change, the agent makes it, and someone needs to confirm the agent actually did what was asked. DraftView closes that loop instead of leaving it open in someone's memory.

Why This Clears the Bottleneck

A queue with a lifecycle changes the reviewer's job from "find the PRs and remember their status" to "work the top of the list."

Triage happens in one place, not across email, Slack, and forty browser tabs.
Status stays explicit, so an approver knows what is waiting on them versus waiting on the agent.
Nothing falls through, because every item has a state and a next step.

The AI handles the volume. The queue handles the order. The humans handle the judgment, which is the only part that was ever theirs to do.

Put your AI doc PRs in one review queue.

Connect a repository and every documentation PR it receives lands in your team's review queue, rendered for reading, with a clear state from new to signed off. Reviewers comment and suggest edits, and everything syncs back to the PR as native GitHub Suggested Changes.

Start your free trial

14-day free trial, no credit card required