The two Markdown calls treat a page as text, so a document can be read, edited and written back without walking the block tree — and the range option means an edit can land in the middle of a long page. Everything structured is still there when the job needs it: data source queries, page properties, comments and the users behind them.
The Notion API as 24 tools: search, users, pages and their properties, blocks and their children, data sources and databases, comments, and two calls that read or write a page's whole content as Markdown.
- A search across the workspace by title, filtered and sorted, and any page, block or database retrieved by ID
- A page's content read as Markdown, and written back as Markdown — inserted, replaced whole, or replaced over a range
- Pages created under a parent with properties, icon and cover, updated, moved, or sent to the trash
- Blocks appended to a page, retrieved with their children, updated and deleted
- Data sources created, retrieved, updated and queried with filters and sorts, and their templates listed
- Comments read on a page or block, and new ones created
- Users listed or retrieved, and the bot user behind the token identified
Npx on your PATH and NOTION_TOKEN in the environment. The integration reaches only what has been shared with it, and the Markdown calls need its read and update content capability — without that they return 403.
One command plus a key — npx @notionhq/notion-mcp-server, then supply credentials