Release Notes

Reader MCP Server with Authentication

The Reader MCP Server now supports authenticated access to private, partial, and password-protected documentation through two authentication types: OAuth 2.1 for interactive client connections, and machine-to-machine credentials for automated access without a browser. Public sites continue to use the existing /_mcp endpoint. Access-controlled sites use the new /_mcp/auth endpoint.

See Reader MCP Server for setup instructions.

Improvements

• MCP tool annotations: All MCP tools now declare standard annotations (readOnlyHint, destructiveHint, idempotentHint) per the MCP specification. MCP clients like Claude and Cursor use these to decide which tools require a confirmation prompt, so read-only tools no longer trigger unnecessary confirmation dialogs.
• Collaborative editing expanded: Real-time collaborative editing now covers Custom CSS settings and Scripts settings editors, not just page content and navigation. OpenAPI files are also handled correctly in the collaboration editor.

Fixes

• Editor going silent after publish: Resolved a critical issue where the editor would stop syncing after publishing, causing pages to disappear from the navigation and subsequent edits to be lost.
• Documentation.json duplication: Fixed documentation.json being duplicated when two clients simultaneously revert or reset the same file, which could corrupt the entire site navigation.
• Images in table cells: Fixed images and videos inside table cells being silently removed on every save.
• HTML table validation: Fixed HTML table elements (table, thead, tbody, tr, td, th) being incorrectly blocked during deployment validation.
• Tabs and CodeGroup underline: Fixed the active tab underline being misaligned when the tab bar is horizontally scrolled.
• BoardCard date props: Fixed BoardCard date attributes not working when written in camelCase (dueDate, createdAt) in MDX. Both camelCase and kebab-case (due-date, created-at) are now accepted.

GitLab Integration

GitLab is now supported as a source repository alongside GitHub. You can connect a GitLab project (on gitlab.com or a self-hosted instance) and use it as the single source of truth for your documentation.

See GitLab Integration for setup instructions.

Connect Hosted OpenAPI Specs by URL

You can now connect an OpenAPI specification hosted at an external URL without uploading or pasting a file. Hosted specs are read-only and cached for fast rendering. See OpenAPI Spec Import for details.

Analytics REST API

The REST API now exposes analytics endpoints for traffic and AI assistant usage. Query traffic totals and daily page views, top pages, referrers, page feedback with written comments, AI assistant questions (including low-confidence filtering), answer ratings, and daily AI question volume. All endpoints accept date range parameters and are read-only, so a viewer API key is sufficient. See API Reference for endpoint details.

Improvements

• Deployments REST API endpoints: The REST API now supports deployment operations, trigger a live deployment, deploy a preview branch, list recent deployments, and poll deployment status from CI/CD pipelines or scripts. See API Reference for endpoint details.
• Reader MCP Server rename: The Docs MCP Server is now called the Reader MCP Server to clearly distinguish it from the Authoring MCP Server. The endpoint, behavior, and existing client configurations are unchanged. See Reader MCP Server.
• Editor bug fixes: Resolved several editor reliability issues, including auto-reconnect after dropped collaboration rooms, publish gating during file operations, file operation failure notifications, and reliable persistence of deleted content.
• Agent bug fixes: Clicking a file link in an AI agent chat result now correctly scrolls to and highlights the specific blocks that were changed in the editor.

Confluence Integration

Connect Confluence as a context source for the AI Documentation Agent. Once connected documentation AI agents can search Confluence spaces and pages, read page content, and use it as context to generate or update documentation

Jira Integration

Connect Jira as a context source for the AI Documentation Agent. The agent can analyze Jira issues and use them as context to automatically generate or update documentation, so your docs stay aligned with what your engineering team is working on.

Improvements

• Collaboration conflict resolution: When publishing from collaboration mode, you'll see a per-file merge dialog if files changed upstream on Git since you started editing. Clean auto-merges are applied silently.
• Agent chat persistence & cancellation: Agent chat history now persists across sessions and devices, and resumes automatically after a refresh. Clicking Stop properly cancels the agent run server-side.

Real-Time Collaboration

Multiple teammates can now edit the same documentation pages simultaneously in the web editor. Changes sync instantly across all collaborators, including page content, navigation, and site configuration.

• Inline comments: Select text and create a comment thread attached to that selection. Reply, edit, delete, resolve, and reopen threads. Comment highlights appear directly in the editor.
• Mentions: Type @ in a comment to mention a teammate from your organization. Mentioned users receive a notification.
• Notifications: A notification bell in the sidebar tracks unread mentions, replies, resolved, and reopened events. New notifications trigger a toast and sound. A daily email digest catches you up on missed activity.
• Collaborator presence: Active collaborators appear as avatars in the editor navbar. Each person's cursor and selection are visible in real time.
• Role-based access: Admins and Editors can edit and publish. Viewers can read and view comments but cannot edit.

Publishing from collaboration mode supports Save, Save and Merge, and Save and Create PR actions. Publishing is serialized per branch to prevent conflicts between simultaneous editors.

Learn more in Real-Time Collaboration.

Improvements

• Faster search: Improved the search algorithm for faster results.
• Ask AI in search: The Ask AI prompt in search results is now always visible, pinned outside the scrollable results list.
• CodeBlock language aliases: node and nodejs now highlight as JavaScript, and curl highlights as Bash. CodeGroup tabs auto-capitalize language names.
• Branch-based editor routing: The editor URL now includes the branch name (for example, /editor/main). Browser back and forward navigation switches branches correctly.

PDF Export

Export your documentation pages as professionally formatted PDF files for offline sharing, compliance archives, or printed handouts. The export runs asynchronously in the background and emails you when the file is ready.

• Trigger a full-site or section export from the dashboard Settings → Advanced → Exports panel.
• Each export is processed in batches, merged into a single PDF, and stored with a 7-day download link.
• Enable Download as PDF as a Page Action so readers can print individual pages directly from the table of contents sidebar. See Page Actions.

SVG and HTML Block Components

Two new content components are now available in the web editor:

• SVG: Embed SVG files with preserved interactivity, including hover states and clickable links. Upload a file, paste SVG code, or select from your existing image gallery via the slash command menu. SVGs render inside a shadow DOM for safe, isolated rendering.
• HTML Block: Insert raw HTML directly into your pages with a toggle between live preview and code editing modes. Useful for custom embeds, interactive widgets, or layouts that go beyond standard markdown.

Both components are accessible via slash commands (/svg and /html) in the visual editor and as JSX components in markdown mode.

SEO Meta Overrides

Pages now support metaTitle and metaDescription frontmatter fields that override SEO metadata, browser tab title, Open Graph tags, Twitter cards, and JSON-LD, independently of the visible page heading and description. This lets you optimize search snippets and social sharing previews without altering what readers see on the page. See Page Configuration.

API Playground Server Management

The API Playground now supports server URL selection and variable substitution from your OpenAPI spec. Readers can pick from multiple defined servers, override variables, and their choice persists across sessions. Parameter inputs with multiple examples show a dropdown picker for quick selection. See Interactive Playground Setup.

AI Agent Architecture;

Dashboard and background agents are now unified, using a streamlined pipeline and can do task more efficiently for the same given credits. Background workflows benefit from improved file operations, PR merge context, and bot commit detection to avoid re-triggering agent-authored changes.

Improvements

• Heading Copy Link&: licking any heading on a published page now copies its anchor link to the clipboard and updates the browser URL hash. A brief check-icon animation confirms the copy, making it easy to share deep links to specific sections.
• Billing Overage Enforcement: Credit deductions are now atomic with overage protection. When you reach your plan limit, deductions and associated side effects are blocked entirely, preventing unexpected overage charges.
• Search Ranking: Page title matching now uses BM25 scoring with a heading-level boost, surfacing canonical top-level pages above same-titled subsections in search results.
• Sidebar Auto-Scroll: The active page in the sidebar now smoothly scrolls into view on first render, so you land on the right context without manual scrolling.
• Sitemap Timestamps: Sitemap lastModified dates now use actual per-page updatedAt timestamps from deployment data instead of the current date, giving search engines accurate freshness signals.
• Danger Zone Settings: Unpublish and Delete documentation actions are now in a dedicated Advanced → Danger Zone settings panel, separated from general settings for clearer organization.
• Navigation Icons: Dropdown menus, tab labels, and group headers in the sidebar and navbar now display configured icons, with active-state color highlighting.
• External Link Auto-Target: External links in your documentation now automatically open in a new browser tab with rel="noopener noreferrer" for security. Internal links and hash anchors are unaffected, and explicit target overrides are respected.

Agent File Attachments

The AI Documentation Agent now supports file attachments. Upload PDFs, code files, documents, and spreadsheets directly into the agent conversation to give it richer context when generating or updating your documentation.

This is especially useful when you need the agent to reference existing specs, design documents, API contracts, or internal guides that aren't already in your documentation project. Instead of copy-pasting content into the chat, attach the source file and let the agent extract what it needs.

Custom CSS

Add custom stylesheets to your documentation site for component-level visual control. Upload .css files or reference external HTTPS stylesheets, then target specific elements using stable class names on 60+ components. See Custom CSS.

Board Component for Roadmaps

A new Board component is now available for organizing content in a kanban-style layout with columns and cards such as Roadmaps

Use it when you want to present information that has distinct stages, categories, or statuses, such as feature roadmaps, project workflows, onboarding checklists, comparison matrices, or migration planning boards. It gives readers a visual overview they can scan at a glance instead of reading through long lists or tables.

Fuzzy Search

Search across your documentation now uses fuzzy matching. Readers can find content even when they misspell terms, use partial words, or don't remember the exact phrasing. This significantly reduces "no results" dead ends and makes self-service discovery more forgiving, especially useful for technical terms, product names, and acronyms that are easy to mistype.

Editor Improvements

• Table of Contents Panel: A new sidebar panel gives you a structural overview of your page, letting you jump to any heading instantly and spot issues like skipped heading levels without scrolling.
• Icon Picker & Color Picker: Visually browse and select icons and colors directly in the editor instead of guessing names or copying hex codes from external tools.
• Drag & Drop Rearrange: Components like CardGroup, Steps, and ParamFieldGroup now support drag-and-drop reordering with editable popovers, so you can reorganize content without cutting and pasting code.
• Turn Into: Convert any block into a different component type (e.g. paragraph → Callout, list → Steps, Card → Expandable) without rewriting content or manually changing markup.
• Editor UI Polish: Smoother transitions, better focus states, improved keyboard navigation, and more consistent spacing across the editor.
• Table Editor Fixes: Resolved cell selection glitches, content loss on rapid edits, and inconsistent column resizing for more reliable table editing.

Visual Change Detection

We’ve added visual change detection so you can review documentation edits in a clearer diff experience before accepting, publishing, or reverting them.

• Review AI changes: Review agent-suggested changes file by file, then accept or reject individual suggestions or all suggestions at once. This also supports AI operations such as create, delete, rename, and move.
• Review saved changes: Compare saved changes on your current branch against the default branch from the Changes view.
• Review unsaved changes: Inspect unpublished edits before publishing, jump back into editing, revert a single page, or revert all pending changes.

Authoring MCP Server

The new Authoring MCP Server is now available as the main MCP experience for read/write documentation workflows. It lets MCP-compatible AI clients work directly with your Documentation.AI project so they can inspect files, update pages, manage branches, review changes, and deploy through your normal documentation workflow.

What you can do with it

• Connect supported AI tools to Documentation.AI using the Authoring MCP endpoint.
• Let agents read and update documentation content, including pages and documentation.json.
• Manage branches and merges as part of AI-assisted documentation workflows.
• Use the same project permissions and role-based access controls that already apply to your documentation workspace.

Learn more in Authoring MCP Server and MCP Server overview.

REST API

We also shipped a new REST API for programmatic documentation authoring and updates. It exposes the same core documentation management capabilities as the Authoring MCP Server over authenticated HTTP endpoints, so teams can integrate documentation workflows directly into scripts, services, and CI/CD pipelines.

What you can do with it

• Read your site configuration and fetch documentation pages over API.
• Search documentation and list available pages programmatically.
• Push documentation updates as Git commits using API keys.
• Create, list, delete, and merge branches through the API.
• Inspect pull requests and automate documentation workflows outside an MCP client.

See API Reference overview for the API entry point.

Analytics improvements

We also shipped several analytics improvements to give teams better visibility into documentation performance and AI usage patterns.

Highlights

• Expanded analytics coverage across traffic, feedback, and Ask AI usage.
• Better insight into how readers engage with your docs, where they struggle, and which content needs improvement.
• Improved Ask AI analytics for reviewing question volume, session activity, confidence signals, and satisfaction trends.
• Easier analysis and reporting with built-in export and dashboard workflows.

See Analytics overview for the full analytics experience.

Features

AI Workflows

AI Workflows let you automate recurring documentation tasks in the background so your docs stay accurate without repeating the same manual work. You can create a workflow from a built-in template or a custom setup, choose how it should run, and review the results from workflow history. Learn more in Workflows.

• Run documentation tasks on a schedule or after pull request merges.
• Use built-in templates for jobs like broken link audits, grammar and typo checks, style guide enforcement, changelog generation, and code-aware documentation updates.
• Add a trigger repository plus additional context repositories for repo-aware workflows.
• Review run history, status, summaries, downloadable reports, and commits generated by completed runs.

Branches and Previews

Create, manage, and deploy preview branches directly from the dashboard. Each branch gets its own live preview deployment so you can review changes safely before publishing. See Branches and previews.

AI Agent Optimization

AI operations are now smarter, faster, and more cost-efficient with compaction, planner-agent architecture, and optimized token management.

AI Assistant Customization

Configure custom instructions, fallback responses, starter questions, style, and widget position to tailor the assistant experience to your docs. See AI Assistant.

Improvements and bug fixes

• AI Assistant Captcha Protection: Added hCaptcha integration to help prevent bot abuse. See AI Assistant.
• Container Content Pages: Navigation containers such as groups, products, and tabs can now have their own content pages instead of only acting as wrappers. See Pages and Groups.
• AI Assistant Follow-up Suggestions: Suggested follow-up questions now appear inline after each answer. See AI Assistant.
• Breadcrumb Navigation: Automatic breadcrumbs now appear on all pages with parent containers. See Pages and Views.
• Documentation Image Quality: Images now render in higher resolution with smarter responsive sizing. See Images.
• Numerous additional improvements and bug fixes across the platform.

CollectionList and CollectionContent components

Two new components are now available to help you build help center-style navigation directly within your documentation pages.

CollectionList renders a grid or list of topic collections, each linking to a dedicated content page. Think of it as the landing page of a typical help center where users see categories like "Getting Started," "Billing," or "Account Settings" laid out as clickable tiles.

CollectionContent displays the articles or entries within a single collection. Use it on a collection detail page to list all related articles with titles, descriptions, and links, giving readers a familiar browse-and-drill-down experience.

When to use them

• You want a help center, knowledge base, or support hub layout where users browse by topic rather than following a fixed sidebar.
• You need a category landing page that groups related articles visually, similar to Zendesk, Intercom, or Freshdesk help centers.
• You want to complement your main documentation with a self-service support section that feels distinct from your API reference or guides.

Together, these components let you create the classic help center pattern, a top-level collection grid that drills down into per-topic article lists, without leaving Documentation.AI or writing custom code.

Pages for containers

Groups and other containers in your navigation can now have their own full pages instead of only acting as folders. This lets you introduce a section, provide context, or add navigation helpers at the top level of a group. Existing structures continue to work, and you can progressively add container pages where they add value. See Pages and Groups for setup details and examples.

JWT and OAuth access control

Access control now supports JWT and OAuth 2.0 so you can protect documentation with the same identity providers and tokens you use in your product. Use JWT to gate docs behind signed tokens you issue from your app, or connect an OAuth 2.0 provider to require sign-in via your existing SSO flow. Both options work alongside existing password protection and integrate with your project-level access settings. Get started in JWT access control and OAuth 2.0 access control.

Cookie consent

Cookie consent tooling gives you a configurable banner and preference controls that help align your docs with privacy requirements. Decide which analytics and third-party scripts require consent and how they behave before and after a user accepts. Configure options in the dashboard and learn more in Cookie Consent.

Integrations

Connect analytics platforms and chat widgets to your documentation with zero-code setup. Built-in support for 17 providers including Google Tag Manager, Google Analytics, PostHog, Amplitude, Mixpanel, Segment, Hotjar, Intercom, Crisp, and more. Just provide your project ID and publish. See Integrations.

Custom scripts

Add custom script files directly to your documentation project as local assets, then reference them from your site configuration to run custom logic. You can also directly add custom scripts directly those from the dashboard. This will be helpful if you want to add any 3rd party integration other that what we support. See Custom Scripts.

Site Config Moved to Editor

Documentation configuration i.e theme, branding, SEO, navigation, integrations, scripts, and more is now accessible directly from the Editor instead of Settings. Click Site Config in the Editor top toolbar. Changes appear in the Changes panel alongside content edits and deploy when you publish. Learn more in Site Configuration.

API Playground enhancements

API Playground now supports multipart/form-data and application/x-www-form-urlencoded bodies in the request builder, including correct cURL and JavaScript code generation and proxy handling for multipart uploads.

Response and code example groups support a dropdown mode for multi-example responses, so you can show multiple variants without overwhelming the page layout.

GitHub Integration UX

GitHub AI context and source repository has a clearer user experience and a more direct navigation path from workspace settings to integrations.

Page Actions

Page Actions add context-aware tools in the table of contents sidebar so you can move content between tools without leaving the reader view. Learn how to configure actions like copy page markdown, view source, open in external LLMs, and copy MCP URLs in Page Actions.

API Playground

You’re no longer limited to generating API docs only from your OpenAPI spec. You now have more control over each API page, for example, you can embed a playground for a specific endpoint from your OpenAPI spec, and then customize the rest of the page content manually; see API Documentation

Page Feedback

Page Feedback now supports custom feedback fields (not only like or dislike) and optional redirects to your GitHub project feedback flow for open source docs. Feedback analytics are richer and roll up into your dashboards; see User Feedback and Analytics overview.

Ask AI Analytics

Ask AI Analytics lets admins review and export the questions users ask in Ask AI so you can see where docs are unclear or missing. This is the first step toward automatically updating documentation from real user intent; explore details in Search and AI assistant analytics and AI Assistant.

Other improvements

• Add helper text to password-gated pages with markdown support, including inline links; see Access Control.
• Enable click-to-zoom on images so readers can inspect diagrams and screenshots in more detail.
• The API Playground now uses OpenAPI example values to prefill parameter inputs so readers can run operations with realistic defaults.
• Many other bug fixes and minor improvements

Access control options

You can now configure access control for your docs at the project level to match how you share content with different audiences. Learn more in the Access Control documentation.

Access levels and example use cases

• Public: Make your entire documentation site visible to anyone.
  • Ideal for fully public product docs and API reference that should be discoverable by customers and search engines.
• Partial: Keep few docs public and few docs private.
  • Gate a few pages for internal teams such as Support, Success, or Sales while keeping the rest of your docs public.
• Private: Require authentication for all documentation.
  • Store internal code documentation, architecture notes, and engineering guides.
  • Centralize internal SOPs, runbooks, and operational checklists for on-call and support teams.

Documentation Templates

Choose from pre-built templates to customize the layout and style of your documentation site.

Available Templates

• Classic - Traditional documentation layout with a clean, focused reading experience. Ideal for technical documentation and API references.
• Atlas - Modern, visually rich layout with enhanced navigation and content presentation. Perfect for product documentation and user guides.

How to use

• Select templates from Settings → General → Brand & Theme in your dashboard
• Configure templates directly in your documentation.json file using the template property
• Further customize your selected template with colors, typography, branding, and navigation settings

Learn more in our Templates documentation.

The default docs domain is migrating from *.documentationai.io to *.documentatioai.com.

• Existing links automatically redirect; no action needed for most users.
• Update references only if you hard‑code the old domain in firewalls, allowlists, scripts, or integrations.
• Custom domains are unaffected and will continue to work as usual.

AI Documentation Agent (beta)

The AI Documentation agent is now available in beta for all workspaces. It helps you generate pages, apply rich visual components, and keep content aligned with your Documentation.AI configuration.

Where can you find it?

• Use the agent from the web editor (using "AI Agent" icon in the editor interface).

**What can you do today? **

• Draft new documentation pages from scratch or from short briefs.
• Make existing docs more visually appealing using rich components (Cards, Columns, Callouts, Tabs, Steps, and more).
• Generate and refine code samples or usage snippets for your APIs and SDKs.
• Use web search inside the agent to bring in up-to-date external information when needed.

Current limitations (beta)

• The agent is currently available only on the web. You can use coding agents like Cursor or GitHub Copilot if you are updating using an IDE.
• The agent understands Documentation.AI’s visual components and documentation structure, but does not yet take context from your codebases or support systems; these integrations are coming soon.
• The agent cannot yet run in the background or autonomously; background workflows are coming soon.
• All AI-generated content should be reviewed and edited before publishing

MCP Server for Your Docs

You can now expose your own documentation/knowledge as a Model Context Protocol server. This makes your published docs directly accessible to your end users, internal teams, and LLM‑powered tools.

Key capabilities

• MCP server for your docs is available by default when you publish your documentation
• Let end users, IDE agents, and chat assistants query your docs through an MCP endpoint you control
• Use structured page and API metadata to give agents more grounded and context‑aware responses
• Help product teams deliver LLM‑native support and troubleshooting experiences directly inside their apps