Eight scoped tools, granted individually
Every token or OAuth connection is scoped to only the tools it needs. A connector that should only propose drafts never gets approve, dismiss, or organize access, and every scope is visible before you grant it.
list_review_inboxSee pending drafts awaiting approval, optionally filtered by project.
create_draftDraft a brand-new article, or propose an update to an existing one.
edit_draftEdit a still-pending draft's title or content before it's reviewed.
approve_draftPublish a pending draft as a new article or article version.
dismiss_draftReject a pending draft without publishing it.
move_article_categoryAssign or clear a published article's category.
reorder_categoriesSet the sort order of a project's article categories.
move_article_projectMove a published article to a different project in the same org.
How it works
From Settings → MCP, create an API token (or connect via OAuth) and choose exactly which of the tools above it can call. A token that only needs to draft doesn't get approve or organize permissions.
Paste your org's MCP server URL into Claude's or Cursor's custom connector settings. The agent can now see your review inbox and help center structure directly from your editor.
An agent can draft, edit, and organize, but nothing it does publishes to your live help center automatically. Every drafted or edited change sits in the Review Inbox until a human approves it.
Every action a connector takes is written to an audit log you can review in Settings → MCP, so you always know which token did what and when. Read access it needs elsewhere (like your GitHub repo, for the docs-drafting pipeline itself) works the same way: scoped, revocable, and read-only where a write isn't the job. See is it safe to give an AI tool repo access for the full breakdown.
Connect your agent, keep the review step
7-day free trial, no credit card required. Create your first scoped MCP token from Settings once you're in.