src/billing/retryPolicy.ts 12 export const retryPolicy = {
13 - maxAttempts: 3,
13 + maxAttempts: 5,
14 + backoff: "exponential",
15 };
DocuWriter.ai Autopilot
Automated documentation that stays in sync with your code.
Connect a Git repository once. Autopilot watches pushes and merged pull requests, finds the documentation affected by each change, and drafts updates your team can review or automatically apply.
Connect GitHub, GitLab, Bitbucket, or Azure DevOps. Start with review mode, then automate on your terms.
Example run
Review mode
Repository change → documentation update
- 00:02 Processing
Push detected
5 files changed on main
- 00:11 Queued
Documentation impact mapped
3 pages need updates
- 00:34 Queued
Suggestions drafted
94% average confidence
- 00:36 Queued
Ready for review
Apply, snooze, or discard
- Files changed
- 5
- Pages affected
- 3
- Avg confidence
- 94%
- Mode
- Review
01 Definition
Direct answer
What is automated documentation?
Automated documentation is a workflow that detects software changes and updates the documentation those changes affect. Unlike a one-off documentation generator, it runs continuously: reading repository diffs, identifying affected pages, drafting edits, and routing them for review or automatic application.
- Reads repository changes
- Pushes, merged pull requests, and manual runs provide a precise change set to evaluate.
- Updates affected pages
- Autopilot targets the documentation connected to the change instead of regenerating every page.
- Preserves editorial control
- Teams can review every suggestion or auto-apply only when confidence and plan eligibility allow.
02 Product evidence
Example update
From code diff to documentation diff.
Autopilot does not need to rewrite an entire knowledge base. It links a specific code change to the pages it affects, explains why, and proposes a focused edit.
- Affected page
- Billing / Retry policy
- Reason
- Retry attempts and backoff strategy changed
- Confidence
- 96% · High
docs/billing/retries.md 8 ## Retry behavior
9 - Failed payments retry up to three times at fixed intervals.
9 + Failed payments retry up to five times using exponential backoff.
10 + Each delay increases before the next attempt.
Review actions: Apply, Snooze, or Discard
Awaiting review Documentation preview
Billing / Retry policy Retry behavior
Failed payments retry up to five times using exponential backoff. Each delay increases before the next attempt.
-
Suggestion ready for review
Confidence scored at 96%. No documentation was changed.
-
Affected page identified
Retry policy documentation matched to the billing source change.
-
Push detected on main
Five changed files entered the Autopilot run.
03 Workflow
Continuous maintenance
Connect once. Keep documentation current after every change.
Define what Autopilot should watch, then let each repository event move through the same traceable maintenance workflow.
- 01
Connect the repository
Choose a Git provider, repository, branch, and the paths Autopilot should include or ignore.
- 02
Detect the change
Start a run from a push, a merged pull request, or a manual trigger when your team wants an on-demand check.
- 03
Draft focused suggestions
Autopilot maps the code diff to affected pages, proposes targeted edits, and attaches a confidence score.
- 04
Choose what ships
Apply, discard, snooze, or bulk-manage suggestions. Eligible plans can auto-apply high-confidence batches and revert them.
04 Controls
Human in the loop
Automation without surrendering editorial control.
Use the same suggestion workflow in two modes. Keep every decision manual, or let a confidence threshold handle routine updates while exceptions remain visible.
Default
Review mode
- Every suggestion enters the review queue.
- Inspect Diff, Document, and History before deciding.
- Apply, discard, snooze, or bulk-manage the queue.
Confidence Any score → review
Eligible plans
Confidence-based auto-apply
- Set an auto-apply threshold for each repository.
- Qualifying suggestions apply; the rest wait for review.
- Audit the batch history and revert an auto-applied batch.
Auto-apply threshold 90%
Example setting. Your team chooses the threshold per repository.
05 Intent map
Choose the right workflow
Automated documentation vs. documentation generation
Documentation generation creates the first version. Automated documentation maintains that work after the code changes. Use the page that matches the job your team needs to complete.
| Your job | Workflow | Best fit |
|---|---|---|
| AI code documentation generator Create documentation from selected source files | Source-file generation | A focused document, API reference, or README |
| automatic code documentation tool Create a documentation set from a full repository | Repository-wide generation | A first structured baseline across a codebase |
| Autopilot automated documentation Keep existing documentation aligned after code changes | Ongoing documentation maintenance | Continuous repository-to-documentation updates This page |
06 Connections
Repository-native
Works where your engineering team already ships.
Configure the repository boundary once. Autopilot uses the events, filters, and notification channels your team already understands.
- Git providers
- GitHub, GitLab, Bitbucket, and Azure DevOps
- Triggers
- Push, merged pull request, and manual run
- Scope controls
- Branch selection plus included and ignored repository paths
- Notifications
- Email, Slack, Microsoft Teams, Google Chat, and custom webhooks
- Confidence controls
- A score on each suggestion and a repository-level auto-apply threshold
- Suggestion actions
- Apply, discard, snooze, bulk-manage, auto-apply, and revert an auto-applied batch
07 Platform proof
Customer story
Repository-scale documentation starts with a reliable foundation.
FinLocker used DocuWriter.ai's generation platform to create documentation across a large repository estate. Their story shows what a repository-scale foundation can look like before ongoing maintenance begins.
Read the FinLocker customer storyWe had over 100 repositories with incomplete documentation. Manually documenting everything would have taken about six months. With DocuWriter.ai and a focused team effort, we were able to complete most of the documentation in just a few weeks.
- Repositories
- 100+
- Platform repositories
- Timeline
- 5 weeks
- Documentation timeline
- Manual path
- ~6 months
- Manual estimate
08
Answers
Automated documentation FAQ
Direct answers about repository monitoring, review, automation, rollback, and plan availability.
What is automated documentation?
Automated documentation is a continuous workflow that detects software changes, identifies the documentation those changes affect, drafts focused updates, and routes them for review or automatic application. It maintains existing documentation after the first version has been created.
How does DocuWriter.ai Autopilot work?
Connect a Git repository and configure its branch and path filters. When a push, merged pull request, or manual run starts, Autopilot reads the change, maps it to affected documentation pages, drafts suggestions with confidence scores, and places them in the review workflow.
Does Autopilot update documentation automatically?
It can, but it does not have to. Review mode lets your team approve each suggestion. On eligible plans, you can set a confidence threshold per repository so qualifying suggestions are applied automatically while lower-confidence suggestions wait for review.
How is Autopilot different from an AI code documentation generator?
A documentation generator creates an initial document or documentation set from source code. Autopilot is the maintenance layer: it watches later repository changes and proposes focused edits to documentation that already exists.
Start with an AI code documentation generator or create a full baseline with the automatic code documentation tool .
Which Git providers and triggers are supported?
Autopilot supports GitHub, GitLab, Bitbucket, and Azure DevOps repositories. Runs can be triggered by a push, a merged pull request, or manually.
Can I review changes before they are applied?
Yes. The review workspace shows Diff, Document, and History views together with the affected page, reason, and confidence score. You can apply, discard, or snooze individual suggestions and bulk-manage a queue.
Can I limit monitoring by branch or path?
Yes. Repository settings let you select a branch and define included or ignored paths so Autopilot monitors only the parts of the codebase relevant to your documentation.
What happens if an auto-applied update is wrong?
Auto-applied suggestions are grouped into batches with an audit history. An eligible user can revert an auto-applied batch and return the affected documentation to its previous state.
Which plans include review, apply, and auto-apply?
Trial and Pro accounts can preview Autopilot suggestions and diffs. Applying suggestions, confidence-based auto-apply, audit history, and unlimited monitoring depend on an eligible Enterprise or Unlimited plan. Check the current pricing page for availability.
DocuWriter.ai Autopilot
Stop letting every merge create another documentation gap.
Connect a repository, review your first suggested update, and decide how much of the maintenance workflow to automate.