My Claude Code now calls my phone when it needs me. No Twilio, no phone number, free
retained text
I kept leaving long Claude Code runs going and coming back to find them stuck on a question for an hour. So I built agentring: an MCP server plus an Android app that let the agent actually call you. In the video a migration fails. Claude Code uses the \`call\` tool and my locked phone rings like a normal call. A voice AI (Gemini Live) tells me what broke and what the options are. I say "backfill them with the legacy plan and run it again". Claude gets the transcript and carries on. \- \`send\_message\` / \`get\_replies\` / \`call\` over MCP: \`claude mcp add agentring -- agentring mcp\` \- Rings even locked and in Doze. A missed call becomes a message you can answer from the notification. \- You can also call it from the app. The voice AI runs \`claude -p\` with your question and reads you the answer. \- Self-hosted with docker compose. No phone numbers, no Twilio, no per-minute fees: it's VoIP from your server to your phone. Android only for now, and so far tested on one phone (Pixel 6). Apache-2.0. The other thing I use it for is agents that watch stuff for me and call when it actually matters: \- a critical vulnerability on Twitter or an official advisory that affects my stack \- a security issue on my servers \- an urgent support ticket or a critical email \- something going down With how many people are out there trying to break into things right now, I think you need agents keeping an eye on your security around the clock. And when they find something bad, a notification isn't enough: your phone has to ring. What would you use it for? Repo: [github.com/mutonby/agentring](http://github.com/mutonby/agentring)
document fetched 2026-10-29 00:42 · canonical https://reddit.com/r/ClaudeCode/comments/1xhvnjw/my_claude_code_now_calls_my_phone_when_it_needs