Advanced Analytica Advanced Analytica: IBOM
← All connectors
Granola logo

Source connector

Granola

Set up Granola as a source and move its data into your chosen destination.

Version 1.1.2 certified generally available

Granola

This page explains how to set up and use the Granola source connector. Granola is an AI-powered meeting notes tool. The connector reads meeting notes from a Granola workspace using the Granola API.

When upgrading across a breaking version, such as 1.0.0, see the migration guide.

Prerequisites

You need a Granola API key from a workspace on a Business or Enterprise plan. Granola provides two kinds of keys:

  • Personal API key: any workspace member can create one. It belongs to that member and inherits their access.
  • Workspace API key: only workspace administrators can create one. It belongs to the workspace, doesn’t expire, and remains active after its creator leaves.

Both keys use the same API endpoints and connector behaviour. The key type determines which notes it can read. See Data access by key type for details.

Set up the connector

Generate an API key

Personal API key

  1. Open the Granola desktop app.
  2. Go to Settings > Connectors > API keys > Create new key.
  3. Select the key’s note access scopes: Personal notes, Public notes, or both.
  4. Click Generate API Key.
  5. Copy the generated API key and store it securely.

:::note On Enterprise plans, a workspace administrator controls which scopes members can use in Settings > Workspace > General > API access for members. Members can’t create keys with scopes disabled there. The connector can’t read notes covered by those scopes. :::

Workspace API key

  1. Open the Granola desktop app as a workspace administrator.
  2. Go to Settings > Connectors > Workspace API keys > Create new key.
  3. Copy the generated API key and store it securely.

Set up Granola in Airbyte

  1. Enter a Name for the Granola source connector.
  2. Enter your API Key.
  3. Optionally, enter a Start Date in YYYY-MM-DD format. The connector replicates notes last updated on or after this date. If you leave this field empty, it replicates notes updated within the last two years.
  4. Click Set up source and wait for the connection test to finish.

Supported sync modes

The Granola source connector supports these sync modes:

FeatureSupported?
Full Refresh SyncYes
Full Refresh Sync - OverwriteYes
Incremental SyncYes
Incremental Sync - AppendYes
Incremental Sync - Append + DedupedYes

Supported streams

The Granola source connector supports these streams:

StreamSync modePrimary key
notesIncrementalid
detailed_notesIncrementalid
note_transcriptsFull refreshNone

Notes

The notes stream retrieves meeting notes from your Granola workspace through the GET /v1/notes endpoint. Each record includes the note ID, title, object type, owner name and email, and creation and last-update timestamps. The API may return other fields, which the connector captures automatically.

For incremental syncs, the connector uses updated_at as the cursor field. It passes the stored cursor in the updated_after query parameter. The API has no upper-bound update filter. Therefore, the connector reads every note updated since the cursor in one request window. The window begins one second before the stored cursor, preventing missed notes updated within that second. If someone edits a note after a sync, the next incremental sync emits it again. In Incremental | Append mode, each edit adds a new row. Each sync also repeats the most recently updated note.

The connector reads data in one window, so it saves the notes cursor only after finishing the stream. If a sync fails during notes, the next sync restarts from the last saved cursor. If no cursor exists, it restarts from your start date.

The API returns only notes with a generated AI summary and transcript. It excludes notes still being processed or never summarised.

The Granola API neither returns deleted notes nor marks them as deleted. Therefore, the connector can’t report deletions. Deleted Granola notes remain in your destination until you run a full refresh with overwrite.

Detailed notes

The detailed_notes stream retrieves each note from the notes stream through the GET /v1/notes/{note_id} endpoint. It includes note metadata and fields available only from the detail endpoint. These fields include summaries, transcripts, attendees, calendar events, and folder membership.

For incremental syncs, the stream uses updated_at as the cursor field. Its state keeps a separate copy of the notes cursor. It uses this cursor to list notes with updated_after. After the first sync, it requests details only for notes updated since the previous sync. Because this copy is separate, syncing, clearing, or refreshing notes doesn’t change which notes detailed_notes requests. The connector always requests transcript data for this stream.

In Incremental | Append mode, each edit adds a row for the note. Each sync also repeats the most recently updated note because notes uses the same one-second lookback.

Incremental sync for detailed_notes is available from version 1.1.0. Existing connections using full refresh continue doing so until you change the stream’s sync mode.

Because detailed_notes uses notes as its parent stream, it requests details only for notes returned by the list endpoint. The connector skips a note without producing a record in two cases:

  • Granola responds with 413 and error code TRANSCRIPT_TOO_LARGE because the transcript is too large to return inline. Long recordings, such as multi-hour meetings, are most likely to reach this limit. Sync the note_transcripts stream to replicate these transcripts. Versions before 0.3.0 failed the sync in this situation.
  • Granola lists the note but responds with 404 NOTE_NOT_FOUND when the connector requests its details. Granola may return 404 for the same note on every sync. Versions 1.1.1 and earlier failed the entire stream in this situation.

Later incremental syncs request a skipped note again only after someone edits it. See Notes are missing from detailed_notes to identify skipped notes.

Note transcripts

The note_transcripts stream retrieves each note’s transcript from the GET /v1/notes/{note_id}/transcript endpoint, with one record per transcript segment. Each record contains the segment’s speaker, text, start and end times, and the note’s note_id. This paged endpoint returns transcripts of any size, including those detailed_notes can’t return inline.

The stream has no primary key, so it appends records instead of deduplicating them. It isn’t incremental. Every sync rereads all notes updated since your start date and requests their transcripts again. It therefore adds at least one request for every note in that range, not only new notes. The connector requests the API’s maximum of 100 transcript segments per page. Transcripts exceeding 100 segments require one extra request for each additional page. With thousands of historical notes, these requests dominate sync time. Set your sync frequency according to the total note count.

Granola may stop returning a transcript for a note listed by the notes stream. This can happen when someone deletes or unshares a note during a sync. The API then responds with 404, and the connector skips that note instead of failing the stream.

Data access by key type

Your configured key determines which notes the connector can read:

Key typeData scope
Personal API keyThe scopes selected when the key was created. Personal notes covers notes you own, notes shared directly with you, and notes in private folders shared with you. Public notes covers notes visible to everyone in the workspace, such as notes in the Team space.
Workspace API keyPublic notes in the workspace, plus notes in spaces where Allow Granola API access is turned on. Granola turns this setting on by default for new spaces. For an existing space, a workspace administrator can open the space, click Integrations in the space header, choose Granola API, and switch on Allow access with a workspace API key. Folders inherit this setting from their space. If an administrator turned off Allow public folders for the workspace, the key can’t read public notes. Private notes and folders that weren’t shared this way are excluded.

Granola notes are private by default. A key with only Public notes access returns nothing until notes enter a folder visible to everyone in the workspace. If a sync returns no records, check the key’s scopes first. For more information, see the Granola API documentation.

Performance

The Granola API enforces rate limits. Depending on the key’s access scope, these limits apply per user or workspace.

MetricValue
Burst capacity25 requests
Time window5 seconds
Sustained rate5 requests per second (300/minute)

The connector limits requests to the documented burst limit of 25 requests per 5 seconds. If Granola still returns 429 Too Many Requests or a 5xx server error, the connector retries up to 5 times. When available, it follows the interval in the Retry-After response header, up to 60 seconds. Otherwise, it uses exponential backoff.

Troubleshooting

Syncs fail with 401 or 403

When Granola returns 401 or 403, the connector immediately fails the sync with a configuration error. It doesn’t retry.

  • 401 Unauthorized means Granola rejected the API key. The key may be mistyped, revoked, or expired. Workspace API keys don’t expire, but personal API keys can. Create a new key in the Granola desktop app, then update the connector’s API Key.
  • 403 Forbidden means Granola accepted the key but denied access to the requested data. Compare the key’s access scopes with Data access by key type.

Notes missing after version 0.2.13 or earlier

Versions up to 0.2.13 sent each 30-day window’s bounds as dates instead of timestamps. The Granola API excludes the entire day named by created_before. Those syncs therefore skipped every note created on a window boundary date. This affected both the notes and detailed_notes streams. Version 0.2.14 sends second-level timestamps, so new syncs cover the full range.

Existing connections don’t automatically backfill skipped notes. After upgrading to 0.2.14 or later, refresh the notes stream once to recover them. If you upgrade directly to 1.0.0 or later, you don’t need a separate refresh. The first sync after upgrading rereads every note updated since your start date. You don’t need to do anything for detailed_notes. Before version 1.1.0, it supported only full refresh. After the fix, each sync reread every note since your start date.

Notes missing from detailed_notes

From version 0.3.0, the connector skips a note in detailed_notes when Granola can’t return its transcript inline. Earlier versions failed the entire sync in this situation. The connector logs each skip at INFO level, not as a warning. Search the sync logs for transcript is too large to identify affected notes. From version 1.1.2, the connector also skips notes when Granola responds with 404. Search the sync logs for returned 404 when its details were requested to identify those notes. A note skipped because of 404 still appears in notes, but may have no transcript in note_transcripts.

A note skipped because its transcript is too large still appears in notes. Its record includes the ID, title, owner, and creation and update times. Its transcript also remains available in note_transcripts. Only detail endpoint fields are unavailable: summaries, attendees, calendar events, and folder membership.

IP allow list

If you use Airbyte Cloud and restrict access to specific IPs, add the Airbyte Cloud IP addresses to your allow list.

Reference

This connector uses the Granola API. Every API request uses the https://public-api.granola.ai endpoint.

For programmatic configuration, use these parameter names:

FieldRequiredDescription
api_keyYesGranola API key. Use a personal API key for notes your own account can read, or a workspace API key for the workspace’s shared notes.
start_dateNoEarliest note update date to replicate, in YYYY-MM-DD format. Defaults to two years before the sync runs.

Changelog

Expand to review
VersionDatePull RequestSubject
1.1.22026-11-1091289Skip notes that Granola returns as not found in the detailed_notes stream instead of failing the sync
1.1.12026-11-1091556Update dependencies
1.1.02026-11-0489049Add incremental sync to the detailed_notes stream
1.0.62026-11-0390584Update dependencies
1.0.52026-10-2789923Update dependencies
1.0.42026-10-2089255Update dependencies
1.0.32026-10-1388655Update dependencies
1.0.22026-10-0687885Update dependencies
1.0.12026-10-0587682Promote to certified
1.0.02026-10-0287011Sync notes incrementally on updated_at so edited notes are replicated. Append-mode notes connections get duplicates on the first sync; see the migration guide
0.3.52026-09-2986915Add Granola-specific messages for 401, 403, 429, and 5xx errors, add heartbeat timeout, suggest notes and detailed_notes, and fix icon dimensions
0.3.42026-09-2987192Update dependencies
0.3.32026-09-2286667Update dependencies
0.3.22026-09-1586082Update dependencies
0.3.12026-09-0885516Update dependencies
0.3.02026-08-2184907Add note_transcripts stream for transcripts of any size. Notes whose transcript is too large to return inline are now skipped in detailed_notes, producing no record for those notes, instead of failing the sync
0.2.142026-08-2184898Stop dropping notes created on a 30-day incremental slice boundary date
0.2.132026-08-1884623Update dependencies
0.2.122026-08-1284278Retry rate-limited and server-error responses with backoff honoring Retry-After
0.2.112026-08-1183964Update dependencies
0.2.102026-08-0483481Update dependencies
0.2.92026-07-2882970Update dependencies
0.2.82026-07-2182437Update dependencies
0.2.72026-07-1481869Update dependencies
0.2.62026-06-3081096Update dependencies
0.2.52026-06-2380491Update dependencies
0.2.42026-06-1679886Update dependencies
0.2.32026-06-0979357Update dependencies
0.2.22026-06-0277288Update dependencies
0.2.12026-05-1578117Update API key setup instructions
0.2.02026-05-0777861Promoted release candidate to GA
0.2.0-rc.42026-05-0177698Revert default_concurrency from 6 to 5 (optimal value from tuning) and add HTTP API budget matching Granola’s documented rate limit (25 req/5s burst)
0.2.0-rc.32026-04-3077645Increase default_concurrency from 5 to 6 for concurrency tuning iteration 3 (final)
0.2.0-rc.22026-04-2877551Increase default_concurrency from 4 to 5 for concurrency tuning iteration 2
0.2.0-rc.12026-04-2777067set default_concurrency=4 for concurrency tuning iteration 1 (Path A, max_rate_limit=5 req/s)
0.1.32026-04-2176632Update dependencies
0.1.22026-03-3175737Update dependencies
0.1.12026-03-2475353Update dependencies
0.1.02026-02-2574033Add detailed_notes substream with full note content via SubstreamPartitionRouter
0.0.32026-02-2473377Update dependencies
0.0.22026-02-1273306Fix pagination: set page_size to API maximum of 30 and improve stop condition
0.0.12026-02-1173238Initial release
Advanced Analytica

Most businesses are sitting on information they can’t use at speed. The people who hold it are busy, and when they leave, the knowledge goes with them. The documents are static, and the processes around them weren’t designed for AI.

Advanced Analytica works with these businesses to turn that information into structured data assets. Once they're integrated, AI systems can govern it, and the business can use it. We bring in the strategy, the specification and the AI skills to rework how the business operates.

MOVE FAST. STAY SAFE.