August 23, 2026 · release

@voicethere/agent 0.4.0 · @voicethere/cli 0.12.0

VoiceThere can now store voice-session audio when you opt in per project — user microphone plus agent TTS — with playback on the session detail page and CLI download for local copies. Your agent decides when capture runs so you can honor consent and skip sensitive audio.

What shipped

Opt-in per project — conversation_recording_enabled defaults to off. Turn it on in Session settings and deploy.
Included minutes — Free 60, Budget 120, Budget+ 240, Advanced 500, Ultimate 1,000 recorded minutes per UTC month (retention follows your plan).
Rollover or metered — With metered recording overage off (default), unused minutes roll into a capped bank at month end. With metered on, unused minutes do not roll over; extra minutes bill at 1 credit/min when billing and a payment method are ready (paid tiers only).
CLI download — voicethere sessions recording <id> --wait --output ./session.opus (or --json for metadata).
Agent controls — await startRecording, pauseRecording, resumeRecording, and stopRecording for consent gates and PII pauses (paused audio is omitted from the file). Check recordingAvailable on session start; local verify resolves with local_mock so laptop runs do not block on a runner parent.
Project-scoped — Recordings, banks, and playback URLs belong to the authenticated project only.

Try it

Docs: Conversation recording
Dashboard: Project overview → Session settings → Conversation recording
Subscriptions: compare included minutes on Plans

Enable recording and deploy:

voicethere projects session-settings set conversation_recording_enabled true
voicethere deploy --wait

Download a recording after a session ends:

voicethere sessions recording <orchestratorSessionId> --wait --output ./session.opus

Only start recording after the caller agrees:

import { defineAgent, speak, startRecording, stopRecording } from "@voicethere/agent";

export default defineAgent({
  async onSessionStart({ sessionId, recordingAvailable }) {
    if (!recordingAvailable) return;
    await speak(
      sessionId,
      "This call may be recorded for quality. Say yes to allow recording, or no to continue without it.",
    );
  },
  async onUserSpeechFinal({ sessionId, text }) {
    const answer = text.trim().toLowerCase();
    if (/\bno\b|do not|don't|decline/.test(answer)) {
      await stopRecording(sessionId);
      await speak(sessionId, "Understood — we will not record this call.");
      return;
    }
    if (/\byes\b|ok|okay|agree|allow/.test(answer)) {
      const result = await startRecording(sessionId);
      if (!result.ok) return;
      await speak(sessionId, "Thanks — recording is on.");
    }
  },
});

Pause while the caller shares sensitive details (card numbers, passwords, and similar) so those frames never land in the recording — see the full pause/resume example in the docs.

For metered overage, add a payment method under Organization → Usage and enable metered recording overage on the project — see Billing & usage credits.

Conversation recording with agent pause and resume

What shipped

Try it

Similar posts