Skip to main content

Call recording and notifications to your CRM

Module goal: turn on recording with a retention period and hand the call results to an external system.
Recording parameters gear: channels and retention

Call recording: stereo (agent and customer in separate channels) or mono; retention is forever or a number of days. Stereo makes reviewing disputed calls much easier.

Recording parameters gear: channels and retention
Recording is turned on with the Call recording switch on the Call card, and the Recording parameters gear next to it opens two settings:
  • Channels — Stereo (agent and customer in separate channels) or Mono (a single mixed channel).
  • Audio retention — Keep forever or Delete after N days, with a period of 1 day or more. This is your answer to the question “how long do we keep the customer’s voice”.
Voice calls are recorded — phone calls and test calls from the browser. The finished recording appears in the player in the header of the call details, where it can also be downloaded; once the retention period is over, the card shows The recording was deleted according to the retention policy.
Advice: tell the caller about the recording. Call recording is regulated by law, and the rules depend on the country — and in the USA, on the state as well: some states require the consent of one party to the conversation, others the consent of all parties. The platform does not warn the caller by itself — that is your organization’s responsibility. A reliable way is to mention the recording in the agent’s first reply and check the wording with your lawyer.
Webhooks window: address, events and the Test button

Post-call notifications: the address, the set of events and the Test button. The address settings have a separate Send test-session events switch — off by default, so tests don't clutter your CRM.

Webhooks window: address, events and the Test button
Two events you can subscribe to:
  • call_ended — the call is over: status, reason, duration, transcript, variables.
  • call_analyzed — post-call analytics is ready: summary, sentiment, custom fields, the success flag, plus the full call data and the finished recording. The event also arrives when the analytics was skipped or failed — the reason is given in the event itself.
Delivery is retried; if a secret is set in the Signing secret — env variable name field, the request is signed, and the receiving side can verify that it is genuine. The deliveries for a call are listed in the Webhook deliveries block of the call details, and undelivered ones can be resent with the Retry button. Webhooks are saved right away — they are not part of the agent draft and don’t need publishing.
An address belongs to one agent, and it pays to know this in advanceA webhook belongs to the agent in whose settings you created it; there is no organization-wide webhook. If you need events for several agents, add the address to each of them — the URL can be reused, and agent_id in the payload tells them apart.
Variables arrive on an unanswered call too. A call that did not connect produces a webhook with dynamic_variables — the values passed for the contact. An empty dictionary is not a sign of an unanswered call: tell the outcome by call_status.
A campaign call carries the campaign and the contact in metadata: campaign_id, contact_id, attempt_no, campaign_name, campaign_created_via and campaign_launched_via. What arrives in the events, with examples, is on the Webhooks page; signing, retries and delivery order are in the Web API.
PracticeGo through all the settings sections of your agent and fill them in as shown: the prompt, two models, a voice with a fallback, timers, AMD, recording, a webhook. Then click Save — and move on to conversation quality.