Call4me Lets AI Agents Make Your Calls for $0.25 a Minute

Nick Khami’s Call4me charges $0.25 a minute to let an AI agent phone a business on your behalf. Ask ChatGPT, Claude Code or Codex to move a dentist appointment, and the agent dials, sits through the phone menu, talks to whoever answers and sends back a transcript.

What happened

Khami launched the service on September 30. It connects to agents through MCP, the standard that lets AI assistants use outside tools. You add one server URL, sign in, and then ask for calls in plain language. Call4me says it works with Claude Code, Codex, Claude Desktop, ChatGPT and Grok, and that Meta’s Muse AI agent can build its own integration.

The idea came from frustration. Khami wanted his agents to book medical appointments and couldn’t find a managed product that did it, so he built one. His write-up of the build explains how it works. There’s just one model on the audio path, OpenAI’s GPT-Live, which hears and speaks phone audio directly. A separate text model makes the decisions, like pressing keypad digits or hanging up. Telnyx carries and records the calls.

The agent also won’t dial until it has what the call needs. Call4me sorts calls into 12 categories, each with its own required details. “The caller can only say what it was given,” Khami wrote. If a business asks something unexpected, the question goes back to your agent, and your answer is spoken on the line.

What this means for students

Students get the best deal. Call4me offers $50 in free credits to students, which it says covers about 200 minutes of calls. Every request is reviewed by hand, and a school email address is enough to apply.

The company says it built the tool with particular students in mind: deaf and hard-of-hearing students, students with speech differences, autistic students and anyone with phone anxiety. Nobody has to explain why or show documentation. The listed uses are deliberately boring. Rescheduling a dentist. Transferring a prescription. Waiting on hold with an airline.

If you’ve been putting off a phone call for weeks, that’s a real offer.

Call4me is upfront about the limits. It isn’t a relay service or certified assistive technology, and it isn’t for emergencies. It also won’t read out card numbers, bank details or passwords.

Pricing, coverage and the fine print

Call4me bills $0.25 per minute from pickup, and phone menus count as talk time, rounded up to the minute. Hold time costs half. Unanswered, busy and failed calls are free. Credits start at $10 and reload monthly by default, so untick that box if you only want to try it. Some destinations cost more, including the UAE and Japan at $0.40 a minute.

The margins look healthy. Khami puts GPT-Live’s cost at about 5¢ a minute, and a Stripe screenshot on his blog shows $610 in gross volume in the first week.

Coverage is less clear. Call4me’s FAQ says its free US number reaches the US, Canada, the EU, the UK, the UAE, Japan and others. ChatGPT’s plugin listing, which spells the name CallForMe, says calls go only to businesses in the US and Canada. If you’re outside North America, test it before you rely on it.

On disclosure, the agent doesn’t open by saying it’s an AI. If someone sincerely asks, it says yes. It calls as your assistant and never claims to be you.

What to watch next

Phone menus are still the weak point. Khami writes that speech menus at FPL and UPS looped for about 8 minutes until the caller learned to say “representative.” Before he added a filler line, one Spectrum rep sat through 41 seconds of silence while the AI waited on an answer.

Businesses are putting AI on their side of the line too. In Call4me’s own tally, calls answered first by an AI assistant took longer to reach a person: a median of 3:07 across 11 calls, against 1:56 across 19 that began with a plain menu. RuntimeWire notes it’s a small first-party sample, not a benchmark.

Call4me sits alongside other experiments in AI agents booking flights and running errands. It hasn’t published total call volume or failure rates.

Watch for those numbers, and for how call centers respond as more of the voices waiting on hold stop being human.