Recording & transcription

Recordings and transcripts are off by default. Enable them per-agent when you need them.

Common reasons to enable:

• QA and coaching. Review calls to evaluate agent performance and train human staff.
• Compliance. Some regulated industries require call recording. Transcripts can satisfy documentation requirements.
• Analytics. Feed transcripts into downstream pipelines to extract intent, sentiment, or named entities.
• Automation. Trigger workflows based on call outcomes by parsing transcript data in webhooks.

Set `recording_enabled` and `transcription_enabled` on the agent via a PATCH request.

curl -X PATCH https://api.aethexai.com/api/v1/agents/$AGENT_ID \
  -H "X-API-Key: $AETHEX_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "recording_enabled": true,
    "transcription_enabled": true
  }'

Recordings are stored as MP3 files. Processing takes a few seconds after the call ends.

List recordings via GET/recordings — filter by `status =` ready to get only finished recordings.

Download the audio file via GET/recordings/{id}/audio, which returns a short-lived presigned URL.

Transcripts are structured as a list of turns with role, text, and timestamp.

Retrieve a transcript via GET/conversations/{id}/transcript.

{
  "conversation_id": "f4a2c8e0-1234-...",
  "turns": [
    {
      "turn_id": "8a91b3c2-...",
      "role": "assistant",
      "turn_index": 0,
      "text": "Hi, thanks for calling First Capital Bank. This is Maya. How can I help?",
      "created_at": "2026-05-12T14:21:03.142Z"
    },
    {
      "turn_id": "12c4d8f6-...",
      "role": "user",
      "turn_index": 1,
      "text": "Hey, I'm trying to figure out how to dispute a charge on my card.",
      "created_at": "2026-05-12T14:21:08.974Z"
    },
    {
      "turn_id": "9d7a6e3b-...",
      "role": "assistant",
      "turn_index": 2,
      "text": "For a card dispute, let me transfer you to a specialist on our fraud team. One moment.",
      "created_at": "2026-05-12T14:21:14.508Z"
    }
  ]
}

The role field is "user" (caller), "assistant" (agent), or, after a conference transfer, "human_agent" (the person the call was transferred to).

Poll GET/recordings until `status` is `ready` for the recording.

For transcripts, poll GET/conversations until `transcription_status` is `completed`.

Recordings and transcripts are retained for 90 days by default. Contact our team if you need custom retention.

Delete a recording early via DELETE/recordings/{id} . This also removes the associated transcript.

For the full request and response shapes, see Recordings and Conversations in the API reference.