Voice (SIP, PSTN)

Saaya runs voice agents over SIP. The agent configuration is the same one you use on chat and video, only the surface changes.

Available numbers

List the numbers available to your organization, then use one as the caller ID when you place a call. Self-serve number purchase and bring-your-own-carrier attachment are not available over the API yet, talk to us if you need a number added.

Outbound calls

Place a call by naming an agent, a number you own to call from, and a number to call. Saaya dials, runs the agent, and writes a session record you can read back afterwards.

curl https://api.saaya.ai/api/v1/phone-numbers \
  -H "X-API-Key: sa_..."

curl -X POST https://api.saaya.ai/api/v1/phone-numbers/call \
  -H "X-API-Key: sa_..." \
  -H "Content-Type: application/json" \
  -d '{
    "agent_id": "…",
    "from_number": "+14155550123",
    "to_number": "+919876543210"
  }'

Call handling

• Speech-to-text runs on the incoming stream while the agent speaks, so the caller can interrupt.
• Turn detection and interruption sensitivity are configured per agent, not per call.
• Voicemail detection can hang up or leave a message instead of talking to an answering machine.
• Keypad input is detected during the call, so a caller can press a digit rather than speak it.
• A maximum call duration can be enforced per agent so a stuck call cannot run indefinitely.

Recording

Recording is configured per agent. Audio-only calls are stored as an OGG file; calls with an avatar or composite video are stored as MP4. Fetch a recording through the session it belongs to.