What the Platform AI Agent can do

The Platform AI Agent helps you search, understand, improve, and manage documentation from the in-app chat panel.

For a quick overview of which actions happen in chat, which need confirmation, and which require opening a setup page, see Agent actions and approvals.

Search your documentation

Ask the assistant to find pages across Spaces you can access. It uses smart search to surface the most relevant results, even if you do not remember the exact title.

Example: "Find all pages about API authentication"

Read documents

Ask to see the full content of a page you have access to. The assistant retrieves and displays it in chat so you can review a page without navigating away.

Example: "Show me the Getting Started page in my API docs"

Browse Spaces

Not sure what is where? Ask the assistant to help list and browse Spaces, folders, and pages it can access. For best results, name the Space when you want to browse a specific documentation tree.

Example: "What pages are in my Platform Docs space?" or "List everything in my API Docs space"

Update existing pages

When you ask the assistant to edit a page, it proposes the changes with a side-by-side preview and color-coded diff:

  • Green highlights show added content
  • Red highlights show removed content

Nothing changes until you review the diff and click Apply Changes. You can also click Reject to discard the proposal.

Example: "Fix the typo in the second paragraph" or "Rewrite the introduction to be more concise"

Create new pages

Ask the assistant to create a new document in a Space where you have edit access. It generates a preview first. The page is only created when you click Create Document.

Example: "Create a Deployment Guide in my API docs space"

Search source code

If an accessible Space is linked to a repository, the assistant can search source code to answer implementation questions or help you document code behavior.

If no matching repository is linked, the assistant can still help with existing documentation, but source-aware checks work best after connecting the matching repo.

Example: "Search for how the user authentication middleware works"

Check documentation against source code

The assistant can help check whether a documentation page still matches linked source code. It can point out likely outdated information, incorrect details, or missing content, then propose updates for you to review.

Example: "Review this page against the linked repository" or "Check whether the API endpoint documentation still matches the code"

Review Autopilot suggestions

If Autopilot has created documentation update suggestions, the assistant can help you find pending suggestions and summarize what each one is about.

When you ask to apply a suggestion, the assistant prepares the request for review. The suggestion is not applied until you confirm it.

Example: "Show pending Autopilot suggestions for this space" or "Apply the authentication suggestion"

View Space activity

When your plan includes activity logs, the assistant can show recent Space activity, such as documentation changes, sync events, and settings updates.

Example: "What changed in this space this week?"

Get help with DocuWriter.ai

Not sure how a feature works? Ask the assistant. It searches the official DocuWriter.ai documentation to find answers and links to the right pages.

Example: "How do I set up the MCP integration?" or "What file types can I import?"

Get guided help for setup and account tasks

For workflows that require an external account, browser download, payment portal, or provider authorization, the assistant can explain the steps and point you to the right page.

Examples include Git provider connections, Confluence setup, exports, webhooks, AI preferences, billing, password changes, and cancellation.

Contact support

Need human help? Ask the assistant to contact support. It drafts a message based on your conversation, and you can review and edit the message before sending. Nothing is sent until you click Send to Support.

Example: "I need help with my billing" or "Contact support about this issue"

Reviewing multiple changes

When the assistant proposes several changes at once, you can navigate through them one by one. The interface shows your position, such as "1 of 3", with navigation arrows.

You have these options:

  1. Review individually: step through each change and approve or reject one at a time
  2. Approve All: accept all pending content and suggestion changes at once
  3. Reject All: discard pending items

Support requests stay separate from Approve All so you always control what gets sent.

Permissions

The assistant respects your existing access permissions:

  • It can search and read documentation your account can access
  • It can edit or create pages only in Spaces where you have manage or edit access
  • It cannot access Spaces that are not shared with you
  • Some actions may require a paid plan, team role, or connected integration