AI Agent

Fetches the full MDX content of a page by path. The agent uses this to understand existing content before proposing changes. Returns the raw MDX source including frontmatter.

Parameters: path (string) — the page path, e.g., /quickstart Returns: { title, description, content, lastModified }

Performs hybrid search (keyword + semantic) across all pages. The agent uses this to find related content, detect duplicates, and understand how topics are covered across the site.

Parameters: query (string) — the search query Returns: Array of { path, title, snippet, score }

Returns the full navigation tree with page titles, paths, groups, and tabs. The agent uses this to understand site structure and identify gaps in documentation coverage.

Parameters: none Returns: { tabs: [{ tab, groups: [{ group, pages: [{ title, path }] }] }] }

Scans pages for internal and external links, then verifies each one. Internal links are checked against the navigation tree. External links are verified with HEAD requests (with a 5-second timeout).

Parameters: pages (string[] | "all") — specific page paths or "all" Returns: Array of { page, link, status, error }

Compares documented API endpoints against the current OpenAPI specification. Detects missing endpoints, changed request/response schemas, updated descriptions, and deprecated operations.

Parameters: specUrl (string, optional) — override the spec URL from docs.json Returns: Array of { endpoint, type, description, severity }

Creates a suggestion with a before/after diff for human review. The suggestion appears in the dashboard and can be approved, edited, or rejected.

Parameters: path (string), description (string), newContent (string) Returns: { suggestionId, status: "pending" }

Commits a change directly to the connected Git repository. Creates a commit on the configured branch with a descriptive commit message. Only available when auto-commit is enabled.

Parameters: path (string), content (string), commitMessage (string) Returns: { commitSha, branch }

Prompt: "Check all internal and external links across the entire documentation site. Report any that are broken or return non-200 status codes."

What the agent does:

1. Calls list_pages to get all page paths
2. Calls check_links with "all" to scan every page
3. Groups results by status (broken, redirected, timeout)
4. Generates a summary with suggested fixes for broken internal links
5. For broken external links, creates suggestions to update or remove them

Prompt: "Compare the API reference documentation against the current OpenAPI spec and update any endpoints that have drifted."

What the agent does:

1. Calls check_openapi_drift to compare docs against the spec
2. For each drift detected, calls read_page to get the current content
3. Generates updated content that matches the spec
4. Creates suggestions (or auto-commits) for each changed page

Prompt: "Find pages that reference deprecated features or outdated version numbers and suggest updates."

What the agent does:

1. Calls search_docs for terms like "deprecated", "legacy", "v1", "old"
2. Reads each flagged page to understand the context
3. Proposes edits to update outdated references
4. Flags pages that may need human review for factual accuracy

Prompt: "Review the documentation structure and identify topics that are missing or underdocumented."

What the agent does:

1. Calls list_pages to get the full navigation tree
2. Searches for common documentation topics (authentication, errors, rate limits, etc.)
3. Compares against the OpenAPI spec to find undocumented endpoints
4. Generates a report of suggested new pages and sections to add