Reference/MCP server
Content tools
These are the MCP tools your agent uses to read and write articles. You rarely call them by name: you describe what you want, and your agent picks the right tool. Knowing what exists helps you phrase requests and understand what happened.
For connection setup, see MCP server overview and authentication. For sections and collections, see Structure tools.
At a glance
| Tool | What it does |
|---|---|
list_projects | Lists projects you have access to |
list_articles | Lists articles, with filters and search |
get_article | Fetches one article in full |
create_article | Creates an article, draft by default |
update_article | Edits an article's draft |
publish_article | Makes an article live |
unpublish_article | Takes an article off the live site |
delete_article | Permanently deletes an article |
get_image_upload_url | Gets an upload URL for an image |
get_directives | Lists supported markdown directives |
Reading
list_projects
Returns every project your account can access, with each project's name, subdomain, and id. Agents usually call this first to find the right project.
list_articles
Lists articles in a project. Supports filtering by collection, by section, by status (draft or published), and text search across titles and content. Returns at most 50 results, so agents narrow with filters on large projects.
get_article
Fetches everything about one article: title, status, collection, full markdown content, meta description, live and preview URLs, any content warnings, and the article's heading anchors. Anchors matter for cross-linking: your agent can link straight to a section of another article with url#anchor instead of guessing.
Writing
create_article
Creates an article in a collection. Articles start as drafts unless the agent passes status: "published" explicitly. Optional extras: a custom URL slug and a meta description for search engines. When an article is published, the response includes the live URL so your agent can hand it straight back to you.
update_article
Edits title, content, slug, or meta description, or moves the article to a different collection. Edits always land on the article's draft. If the article is already live, the published version stays untouched until the next publish. See How authoring works for the draft model.
publish_article and unpublish_article
publish_article promotes the draft to the live site and returns the public URL. unpublish_article removes the article from the live site while preserving the content as a draft, ready to republish later.
delete_article
Permanently deletes an article and its history. There is no undo.
Deletion is immediate and irreversible. If you might want the article back, ask your agent to unpublish it instead.
Images
get_image_upload_url
Hands the agent a short-lived upload URL plus the final public URL of the image. The agent uploads your screenshot or diagram, then references it in markdown as . Accepted formats are PNG, JPEG, GIF, and WebP, up to 10 MB. The upload URL expires after 5 minutes, so agents request one right before uploading. More in Add images and captions.
Discovering formatting
get_directives
Returns the full registry of markdown directives the docs site can render: callouts, tabs, accordions, video embeds, and alignment, with every attribute and allowed value. Agents call this to write valid rich content without guessing. The human-readable version is the Directives reference.
Prompts to try
Find every article that mentions "webhooks" and update them to use the new endpoint name.
Draft an article about our refund policy in the Billing collection, but do not publish it yet.
Take the screenshot at ~/Desktop/settings.png, upload it, and add it to the article about project settings.
