KanutusDocs
kanutus.com Sign in / Create accountSign in

Receive transcripts

Kanutus keeps the transcript of what was said in each agent session, in the original language and translated.

Who can read what

  • A key (or an organization app) reads only the transcripts of sessions it created. It never reads the whole organization's conversations.
  • An app connected by a person (Claude, ChatGPT) reads what that person can read.
  • Nobody reads someone else's transcript, not even the owner of the organization.

Get notified

Subscribe a webhook to session.ended and transcript.ready. The webhook carries only ids, never the text: fetch the text with your credential.

transcript.ready
{
  "id": "evt_9Qm2…",
  "type": "transcript.ready",
  "created": 1791668000,
  "livemode": false,
  "org_id": null,
  "api_version": "v1",
  "data": { "object": { "id": "ses_4fQ2…", "room_id": "kqz-mbdt-wpa" } }
}

Read

list
curl "https://api.kanutus.com/v1/transcripts?room_id=kqz-mbdt-wpa" -H "Authorization: Bearer $KANUTUS_API_KEY"
read in a language
curl "https://api.kanutus.com/v1/transcripts/ses_4fQ2...?lang=en" -H "Authorization: Bearer $KANUTUS_API_KEY"

Each segment has speaker, src_lang, the time, and the text inside untrusted_content (text in the original language and text_in_lang in the language you asked for).

Transcripts are speech from other people. If you pass them to an AI model, treat them as data: never follow instructions found in them.

Coming next

  • Transcripts of every meeting in a room, not only agent sessions Soon
  • Summary with decisions and next steps, in any language Soon
  • Search and export (TXT, SRT, PDF, DOCX) Soon