Coach an agent while the call is still going, without the customer knowing you are there. With Sautikit call supervision, a team lead joins a live call between a customer and one of your agents and picks how involved to be:
The supervisor moves between the three at any point without hanging up or redialling, so the same line serves silent QA, live coaching and escalation.
Supervisors join from wherever they already work: a browser tab in your own app, a desk phone or softphone on a SIP account, or their mobile number.
What it costs: a team lead monitoring from a browser for two hours a day costs about KES 60 a day, at KES 0.50 a minute.
Supervision is part of the Sautikit Voice API, so it works on the calls your agents already take on Sautikit numbers, whether they answer in a browser (including Sautikit Desk), on a SIP phone or in a mobile app. You start and switch supervision through the API, from your own software.
Your server names the call and the supervisor. The supervisor's browser, SIP phone or mobile rings, and they are on the call when they answer.
import { SautikitClient } from "@sautikit/node";
const sautikit = new SautikitClient({ apiKey: process.env.SAUTIKIT_API_KEY });
await sautikit.calls.supervise(callId, {
mode: "listen",
supervisor: { type: "client", identity: "mary" }, // or { type: "phone", number: "+2547..." }
label: "Mary Wanjiku",
});In a browser app you build, @sautikit/webrtc answers the supervisor's leg automatically, so a Listen button in your own interface puts the team lead on the call a moment later.
// Once the supervisor has answered (the call.supervision.started webhook):
await sautikit.calls.setSupervisionMode(callId, "whisper"); // coach the agent
await sautikit.calls.setSupervisionMode(callId, "barge"); // join the conversation
await sautikit.calls.stopSupervision(callId); // leave quietlyModes change only through the API. A supervisor pressing keys on their phone cannot switch into barge by accident.
You decide whether anyone knows a supervisor has joined. Play a tone to the agent, to the agent and the customer, or to nobody. Agents on @sautikit/webrtc can also see a banner such as "Mary Wanjiku is listening", driven by a supervision event in their browser.
Every step is also a webhook, from call.supervision.started to call.supervision.ended with the billable seconds, so supervision time lands in your own QA dashboards and reports.
| Mode | Supervisor hears | Agent hears supervisor | Customer hears supervisor |
|---|---|---|---|
| Listen | both sides | no | no |
| Whisper | both sides | yes | no |
| Barge | both sides | yes | yes |
Recording. Supervision never starts a recording of its own. On a recorded call, whatever the customer heard is in the recording: barge audio is kept, listen and whisper are not. The join tone is never recorded.
Rules. One supervisor per call. The supervisor does not follow a transfer: a transfer ends the supervision, and you can start a new one on the call. AI-agent calls, conference rooms and callers still waiting in a queue cannot be supervised.
A supervisor is billed like any call leg, from the moment they answer until they leave:
A supervisor who does not answer costs nothing. See /pricing for every rate.
Call barging is when a supervisor joins a live call so that both the agent and the customer can hear them, usually to take over a difficult conversation. On Sautikit it is the barge mode, and a supervisor can move into it from listen or whisper without redialling.
Call whisper lets a supervisor talk to the agent during a live call while the customer hears nothing. It is how team leads coach agents in the moment, for example with the answer to a question or a nudge towards the next step.
Only if you choose to tell them. The join tone can go to nobody, to the agent only, or to the agent and the customer. Follow the call-recording and monitoring notices that apply to your business.
Yes. Name the supervisor by phone number and Sautikit rings their mobile. That leg is billed at your normal outbound rate. From a browser or a SIP phone it is KES 0.50 a minute.