Skip to main content
Connect Indexbrain to Notion to search and read your workspace documentation — including structured queries against Notion databases.

​
Setup

1

Go to Integrations

Navigate to Settings > Integrations > Notion.
2

Authorize

Click Connect and select which pages and databases to share.
When connecting, you choose which Notion pages Indexbrain can access — this is Notion’s own sharing screen, shown during authorization. You can update what’s shared later from Notion’s own settings, without reconnecting.

​
Capabilities

Indexbrain can search pages, read their content, and run structured queries against your Notion databases.

​
Search and answer questions

Search results draw on both distilled knowledge facts and, when needed, the raw page text itself. To search Notion specifically rather than every connected source, ask to search Notion directly — this is scoped under the hood so it doesn’t pull in Slack, Linear, or anything else.

​
Query databases

Database rows are pages whose data mostly lives in their properties (Status, Assignee, Priority, Due Date, etc.) — Indexbrain reads all of that, not just page titles, so a question like:
resolves the database by name, builds a real filter against the matching property, and returns the actual rows — not a fuzzy text-search guess.
Database queries currently support one filter condition at a time (e.g. Status = “In Progress”). Combined conditions (“high priority AND assigned to Sarah”) aren’t supported yet — ask one filter at a time, or ask for everything and narrow down in conversation.
Supported filter types: Status, Select, Multi-select, Checkbox. Person-based filters (“assigned to Sarah”) and date-range filters aren’t supported yet — those return unfiltered results with a note explaining why, rather than failing silently.

​
Find decisions in meeting notes

Decisions extracted from Notion pages are tagged as such, so a question like “what did we decide about X” can surface the actual decision along with what it replaced, if anything.

​
How data moves

  1. Live sync — when Notion reports a page was created, edited, or deleted (via a webhook), Indexbrain picks it up and re-indexes just that page within about a minute — no need to wait for a full resync.
  2. Scheduled sync — separately, a full sync runs on your configured processing interval, re-checking everything as a backstop.
  3. Once ingested, a page becomes searchable knowledge within roughly another minute — extraction runs continuously in the background, not on a long batch delay.
Live sync requires a webhook to be registered with Notion — this is a one-time setup step in Notion’s own integration settings, separate from connecting the integration itself. Until it’s registered, Notion falls back to the scheduled sync only.

​
What gets read

  • Page title, body content (up to 4,000 characters per page, nested blocks up to 3 levels deep).
  • All page properties — Status, Assignee, Priority, dates, tags, checkboxes, and more.
  • Inline databases embedded in a page (flattened into a readable table).
  • Top-level databases — their title and property schema are recorded even before a specific row is asked about.

​
What Indexbrain does not do in Notion (yet)

  • No write access. Indexbrain cannot create or update pages, database entries, or tasks — every connected integration is read-only by design.
  • No multi-condition database filters — one property filter per query.
  • No person-name or date-range filters on database queries.

​
Troubleshooting

  • If live sync isn’t registered yet, it can take up to your configured processing interval (default 4 hours) to appear.
  • Confirm the page is actually shared with the Indexbrain integration in Notion (check Notion’s own sharing settings for that page).
  • Very short or empty pages are filtered out before reaching the AI.
  • Confirm the database name you used roughly matches its title in Notion.
  • Confirm the property name and value you’re filtering on exist and are spelled the way they appear in Notion.
  • Remember only one filter condition is supported at a time.