Notion
Service domainDOCUMENTS
Arcade Optimized
Arcade.dev LLM tools for Notion
10tools
The Notion toolkit integrates Arcade with the Notion API, enabling LLMs to read, write, navigate, and search a user's Notion workspace.
Capabilities
- Content access & editing: Retrieve page content by ID or title as Markdown, append Markdown content to the end of existing pages, and create new pages under a named parent.
- Metadata & identity lookup: Fetch full object metadata (ID, timestamps, properties, URL) for any page or database by title or ID; resolve the authenticated user's identity and workspace context.
- Workspace navigation: Retrieve a recency-sampled tree of up to 100 pages and databases, list direct child pages and databases of any object (with cursor-based pagination), and resolve the full breadcrumb path for any single page or database.
- Search: Search pages, databases, or both by title across the workspace (content is not searched).
OAuth
This toolkit uses OAuth 2.0 via the Notion provider. Arcade handles the authorization flow and token management automatically. Users authorize the integration through Notion's standard OAuth consent screen, which grants the integration access scoped to their workspace.
Available tools(10)
10 of 10 tools
Operations
Behavior
| Tool name | Description | Secrets | |
|---|---|---|---|
Append markdown content to the end of a Notion page by its ID or title | |||
Create a new Notion page by the title of the new page's parent. | |||
Find the breadcrumb path for one page or database.
A `complete` path starts at a workspace root. An incomplete path is still the real prefix
of the object's location: `stop_reason` says what ended the walk, and for a rate limit
`retry_after_seconds` carries Notion's own backoff. | |||
Get the metadata of a Notion object (page or database) from its title or ID.
One of `object_title` or `object_id` MUST be provided, but both cannot be provided.
Prefer `object_id` for exact lookup; use `object_title` (with optional `object_type`)
when only a name is known. The title is case-insensitive and outer whitespace is ignored.
Returns all exact title matches in a bounded search window, or a one-item list by ID.
Object metadata includes its ID, timestamps, properties, URL, and more. | |||
Get the content of a Notion page as markdown with the page's ID | |||
Get the content of a Notion page as markdown with the page's title | |||
Get up to 100 recently edited pages and databases as a tree.
The 100 counts pages and databases only: database rows are skipped (they are not sidebar
structure) and reported as omitted_database_rows. This is a recency-limited sample, not a
complete listing: a specific object you are looking for may not appear here at all — search
for it by name and get its location instead. A truncated response means more workspace
objects exist. When sampled_objects is 0 while omitted_database_rows is high, the workspace
is busy with database row edits rather than empty — search by title instead of concluding
the workspace has no content. Unplaced objects are not transitively reachable from a
workspace root in this sample — for example block-nested objects, objects whose parent was
omitted from the sample, or descendants of those. They are not attached under a possibly
incorrect workspace root. Navigate an unplaced object by passing its own `id` (not
`parent_id` or `containing_block_id`) to GetObjectLocation or ListChildObjects. | |||
List direct child pages and databases without counting an entire subtree.
Page children of the requested object come first, then children nested inside its
containers (toggles, columns, callouts). Continue by passing `next_cursor` back; when
`next_cursor` is null there is nothing further to fetch.
`truncated` means the walk stopped before enumerating everything, with `stop_reason` saying
why. If `next_cursor` is also set, continue from it. If `next_cursor` is null, more windows
cannot recover what was missed — every call re-walks the same bounded budget — so do not
keep paging: call this tool again on a narrower parent (a specific child page), or use
GetObjectLocation for one object you already know about. For a rate limit, wait
`retry_after_seconds` before retrying. | |||
Search for similar titles of pages, databases, or both within the user's workspace.
Does not include content. | |||
Get information about the current user and their Notion workspace.
This tool provides detailed information about the authenticated user's
Notion workspace including workspace statistics, user context, and
integration details. |
Last updated on