Notion KB Integration

Credentials required

Step 1 — Create an Internal Integration

1. 1
   Open Integrations

   Go to https://www.notion.so/my-integrations and click New integration.
2. 2
   Configure capabilities

   Set the following capabilities:

   • Read content — required to fetch page blocks.
   • No user information — TR-AI does not need user data.
3. 3
   Copy the token

   Go to the Secrets tab and copy the Internal Integration Token.

Step 2 — Share pages with the integration

The integration only has access to pages you explicitly share with it. For each page or database you want indexed:

1. 1
   Open the page

   Open the Notion page or database you want TR-AI to index.
2. 2
   Share with integration

   Click the ••• menu → Add connections → search for your integration name → click Confirm.

Step 3 — Find your Database IDs (optional)

If you want to index specific databases only (rather than the entire workspace), find each database ID in the URL:

https://www.notion.so/myworkspace/My-KB-a1b2c3d4e5f6...?v=...
                                   ^^^^^^^^^^^^^^^^^^^^^^^^
                                   This is the Database ID (32 hex chars, no dashes)

Enter multiple database IDs as a comma-separated list in TR-AI. Leave blank to index all pages the integration can access.

Step 4 — Enter credentials in TR-AI

In your TR-AI Dashboard, open the Knowledge Base section and select Notion. Enter your Integration Token and optionally your Database IDs. Click Verify & Index. TR-AI will:

• Call POST /v1/search with page_size=1 to verify the token.
• If Database IDs are provided: query each database via POST /v1/databases/{id}/query.
• If no Database IDs: search the entire workspace via POST /v1/search with filter.value=page.
• Fetch block content for each page recursively (depth up to 5 levels).

How TR-AI extracts content

Block types supported

TR-AI walks the block tree recursively and extracts text from:

• paragraph, heading_1/2/3
• bulleted_list_item, numbered_list_item, to_do, toggle
• quote, callout, code
• table_row — cells joined with |
• child_page — recursed into (up to depth 5)

Blocks with has_children=true are recursed automatically. Image, video, file, embed, and divider blocks are skipped.

Pagination

Block children are fetched with page_size=100 and follow next_cursor until has_more=false. Page discovery also paginates exhaustively — no hard cap.

Concurrency

Page content is fetched with concurrency 6 to stay within Notion's rate limits.

Credential verification errors

Notion API version

TR-AI sends Notion-Version: 2022-06-28 on all API calls.