Custom Caller ID on Call Transfers
Control the caller ID shown on warm transfers: use the original caller’s number, a fixed number, or an inbound SIP header.
See Caller ID on warm transfers for details.
Control the caller ID shown on warm transfers: use the original caller’s number, a fixed number, or an inbound SIP header.
See Caller ID on warm transfers for details.

You can now rename conversations in Aurora. Give each thread a clear, descriptive name to keep your work organized.
WebSocket media is now the recommended way to build custom voice integrations for websites and apps. Instead of using the legacy PCM websocket, start a call from your backend and pass the returned WebSocket URL to your client to stream audio.
See the WebSocket media integration guide for details.
Hebrew joins the list of available languages for voice agents.

You can now set the sidebar to collapse, expand, or expand when hovering with the mouse. Find the control in the bottom left corner of the interface.

You can now listen in to a call while it is in progress. Open Call logs, find a call with the in-progress status, and monitor the live conversation. Monitoring is silent, so neither the agent nor the caller can hear you.
See Live call monitoring for details.
We added six new outbound IP addresses for the Global region. If your network uses a firewall allow-list, add these so platform traffic keeps flowing:
34.73.34.13834.26.71.17134.148.30.1034.138.149.22434.139.16.7934.23.185.86See the full list on the Egress IP addresses page. If you don’t restrict inbound traffic, no action is needed.

Upgrade your Voice Widgets to the new design and faster real-time backend. Functionality stays the same.
GPT-4o will be removed from the UI and API on July 9, 2026. On that date, any agents still using GPT-4o will be automatically migrated to GPT-5.2.
This deprecation only applies to non-enterprise accounts.
No action is required.

The agent testing interface has been updated, and the Widget test is now Web Call. Functionality stays the same, so you can keep testing your agent straight from the browser.
You can now use Synthflow directly from Claude Code or Codex via the Synthflow MCP server, giving you access to Synthflow’s core building blocks and operations without leaving your AI workspace.
See the Claude Code and Codex setup guides for installation steps.
Agents can now use shared memory across voice, SMS, and WhatsApp, helping them continue conversations with relevant history from any channel.
You can now connect RingCX to Synthflow over SIP, running your agent as an Intelligent Virtual Agent (IVA) in RingCX voice queues. Route voice queue calls to your agent, pass caller context through Workflow Studio, and hand calls back to the RingCX workflow when the session ends.