Notion

notion wraps the official Notion REST API using an Internal Integration bot token — no OAuth dance. One instance = one integration; it sees only pages and databases that have been explicitly shared to it.

fetch is the MCP-style single-call read: it flattens a page's properties and walks the block tree into clean markdown, matching the shape of Notion's own MCP server.


Source	`plugins/connector/notion/`
Key	`notion`
Icon	📝
Tier	plugin — install with `<app> plugin install notion`
Default tags	`Connector`, `Productivity`

See Connector Plugins for the install flow.

Configs

Field	Type	Required	Notes
`Token`	secret	✅	Notion Internal Integration Secret (starts with `ntn_`). Create one at notion.so/my-integrations.
`Status`	widget		Live connection status card — probes the token against `GET /v1/users/me` and shows the bot name + workspace.

Sharing pages. The REST API only sees what's explicitly shared to the integration. For any page or database the connector needs, open it in Notion and add the integration via that page's Connections menu. A page not shared returns 404, not 403 — that's Notion's permission model, not a connector bug.

All requests use API version 2022-06-28.

Operations

Read

Op	Input	What it does
`search`	`query`, `object_type`, `page_size`	Search pages/databases shared to the integration by title. Empty query lists everything shared.
`fetch`	`id`, `with_content`, `with_blocks`	Fetch one page or database by ID. For a page: flattened properties, plus (default on) the body rendered as markdown by walking the block tree. Set `with_blocks` on to also get `blocks[] = {id, type, text}` for targeting a specific block (e.g. to comment on it). For a database: title + normalized property schema.
`query_database`	`database_id`, `filter`, `sorts`, `page_size`, `limit`	Query a database's rows with an optional raw-JSON Notion filter/sort. Returns normalized rows `{id, url, title, properties}`. Pagination is followed automatically up to `limit`.
`get_comments`	`block_id`	List comments on a page or block.
`get_users`	`query`	List workspace users (people + bots), optionally filtered by name/email substring.

Write

Op	Input	What it does
`create_page`	`parent_type`, `parent_id`, `title`, `properties`, `content`	Create a page — a row in a database or a subpage under a page. `properties` is raw JSON keyed by property name (Notion property-object form); `content` is markdown converted to blocks. Returns `{id, url}`.
`update_page`	`page_id`, `properties`, `append_md`, `archive`	Update a page's properties and/or append markdown to its body. `archive=true` moves it to trash and ignores the other fields. Returns `{id, url}`.
`create_comment`	`text`, `page_id`, `block_id`, `discussion_id`	Add a comment. Target precedence: `discussion_id` (reply into a thread) > `block_id` (comment on a specific block) > `page_id` (page-level).
`create_database`	`parent_page_id`, `title`, `schema`	Create a database under a page from a raw-JSON property schema (must include exactly one title property).
`update_data_source`	`database_id`, `title`, `properties`	Change a database's schema and/or title: add a property (new key), remove one (`null`), or rename one (`{"name":"New"}`).

All write ops are destructive and disabled/enabled per row like any other connector — see Connector Module ▶ Destructive ops.

`fetch` output shape

fetch is built to be the one call an agent needs for "read this page":

meta — flattened {id, url, title, properties, edited_at} (property values normalized to plain JSON, not Notion's verbose property-object form).
content_md — the page body rendered as markdown (default on; turn off with with_content=false to save an API call when only properties are needed).
blocks — flat {id, type, text} list, off by default; turn on with with_blocks=true when you need a block ID to target with create_comment or a future block-level edit.

Notion ​

Configs ​

Operations ​

Read ​

Write ​

fetch output shape ​