Build a terminal calling tool with human backup

They wired a CLI agent to place calls and loop in a person only when it needed an answer.

Done with Custom agent· Advanced· Sep 2026

What they did
They built a command-line phone caller with Twilio and GPT-live so a terminal agent could speak on outgoing calls. The agent stayed on the line by itself, and when it hit something it could not answer, a dashboard panel asked the person for help and sent the reply back into the same call.
How it went
The setup worked in a live walkthrough showing the agent making calls and handing off unknown questions to the person. It was presented as a functioning calling flow rather than a finished product with broader evaluation.
Tips
Keep a human-in-the-loop path for anything the agent cannot safely decide. Test the call flow in dry run mode before letting it dial a real number or say anything sensitive.

Try it yourself with Claude Code

In [project folder], create a small command-line tool that uses Twilio so a terminal agent can place an outbound phone call and speak a short script about [call purpose]. While the call is active, show a local dashboard that displays any questions the agent cannot answer and lets me type a reply that is sent back into the same call without disconnecting. Add a dry-run option that prints the planned call steps instead of dialing, and wait for my approval before any real call is placed, any message is sent, or any recording is saved. Include basic instructions for running a test call and reviewing the outcome.

Fill in the [brackets]. Agents can make mistakes, so review anything before it is sent, paid or deleted.

Read the original on x.com ↗

Get 3 agent uses like this every morning

With the day's top AI news. Free, and one click to unsubscribe.