Call4me lets Claude Code and Codex make phone calls for $0.25 a minute
Founder Nick Khami built the MCP-connected service for tasks such as booking appointments and changing flights; calls are recorded and charged by talk time.
By Ryan Merket · Published
Primary source: X
Why it matters
Call4me extends MCP-connected agents from software tasks to phone calls with businesses, where a scheduling mistake or misunderstood commitment can have consequences beyond a failed API call.
Nick Khami (@skeptrune) introduced Call4me on September 30th, a phone-calling service that lets AI agents including Claude Code and Codex call businesses to handle errands such as booking appointments, changing flights and canceling subscriptions. The service charges $0.25 per minute of talk time, with a $10 minimum prepaid credit purchase, according to Call4me's site.
https://x.com/skeptrune/status/2105320262938009690
Khami's pitch is an interface choice as much as a voice-AI product: he spends much of his workday in a terminal and wanted phone tasks available there instead of forcing a switch to another app. In his launch thread on X, he said he had already used Call4me for dozens of calls. The service connects to compatible agents through the Model Context Protocol, or MCP, so users can ask an agent to call a business in ordinary language and receive the result back in the same workflow.
Khami comes to the launch after founding Trieve, a developer search and retrieval company. He said in Trieve's acquisition announcement that Mintlify acquired the company in July 2025 and that he and co-founder Denzell Ford would join Mintlify. Trieve had raised $3.5 million in September 2024, according to its company history; that funding was for Trieve, and there is no basis to treat it as financing for Call4me. Khami's personal site says he studied computer science at the University of Texas at Austin and was an engineering manager at Mintlify after the acquisition.
Call4me's live product materials describe connections for Claude Code, Codex, Claude Desktop, ChatGPT and other MCP-compatible agents. Users add the MCP server, authorize access and give the agent a request. Calls use a Call4me number, and the agent reports back with a summary. Call4me says it can navigate phone menus and hold queues, and can handle reservations, appointment scheduling, service questions, quotes and cancellation requests.
The price is simple to state but changes the calculation for longer errands: $0.25 per minute, prepaid, starting with $10 in credits. The service says it reserves the call's maximum estimated cost up front and returns the unused balance afterward; unanswered, busy and failed calls are free. Users receive a U.S. number for the first call. Extra numbers are charged at carrier cost, including an upfront charge and monthly renewal, according to the product's rules.
The published examples show both successful handoffs and the limits of treating a completed call as a completed task. In a dentist call dated September 30th, the agent negotiated a Monday appointment with the same provider and the office confirmed the change. A Costco tire-center call found that an order had arrived, but no installation appointment was booked. In a junk-removal example, the agent booked a pickup after receiving a quote, repeated the street number incorrectly, then made a second call to correct it. An Amazon Pharmacy call got a representative to mark delayed prescriptions as important, but the transcript did not confirm that the review was completed. The examples and recordings are published on Call4me's site.
Those outcomes put the product's value and risk in the same place: an agent can take on the waiting and conversational work, but the user still needs to check what the business actually agreed to. That matters when the call changes an appointment, places an order or commits to a booking. Call4me's recordings and transcripts give users a way to review conversations, but they do not guarantee that the agent understood every detail or that a business carried out what it promised.
Call4me says calls are recorded by its telephony provider, Telnyx, with recordings and transcripts available to users. Its rules say the agent does not open by announcing that it is AI, but identifies itself truthfully if asked; it calls as the user's assistant and does not claim to be the user. The service bars telemarketing, surveys, emergency and premium-rate calls, as well as harassment, threats and impersonation.
The launch makes phone calls another tool an agent can invoke, alongside actions in software. It also extends agent automation into conversations where menus, hold times and human representatives still mediate the task. Khami's stated reason for building it is practical: keep the phone call inside the terminal workflow he already uses. Whether that convenience holds up will depend on the accuracy of each call and on users checking the commitments agents make on their behalf.