For AI agents: the complete documentation index is available at https://docs.flezise.com/llms.txt, the full documentation bundle is available at https://docs.flezise.com/llms-full.txt, and this page is available as Markdown at https://docs.flezise.com/integrations/meeting-capture.md.

Live meeting capture

Meeting Capture is an integration that accepts live caption batches from an ongoing meeting and assembles them into a transcript when the session ends. It is gated behind a feature flag; the provider appears locked in the catalog until the flag is enabled on your installation.

Connect Meeting Capture

Open Project SettingsIntegrations and select Meeting Capture. Because Meeting Capture is a receiver, not a remote service, it has no credentials to enter. Press Connect with token to open the connection dialog, then press Test connection. The test passes immediately and the connection is saved.

On save, Flezise generates a capture token and displays it once in the connection panel. Copy it before closing the panel. The token cannot be retrieved afterward; delete the connection and create a new one to rotate it.

Sending captions

Once connected, your meeting bot or capture client sends caption batches to:

POST /meeting/{connection-id}
x-capture-token: <your token>

The request body carries a session_id that groups related batches into one meeting, an array of captions strings, and an optional title. Call this endpoint repeatedly as new captions arrive. Each call appends the captions to the in-progress session and returns {"ok": true, "buffered": N} where N is the count of captions buffered so far.

The endpoint rejects requests with an incorrect or missing x-capture-token with 401 Unauthorized. It also rejects connections that do not have the Live Capture capability with 400 Bad Request.

Finalizing the session

When the meeting ends, call:

POST /meeting/{connection-id}/finalize
x-capture-token: <your token>

The request body carries the same session_id and an optional title. Flezise assembles all buffered captions for that session into a transcript and processes it for storage. The response returns {"ok": true} along with the result details.

A session that is never finalized remains buffered and is not added to the knowledge base.

Next steps