AI agents can write code, manage workflows, and browse the internet on their own now. But ask one to call a customer, send a text, or handle a verification code, and everything suddenly breaks apart. That problem pushed Meet Modi and Manav Modi to start AgentPhone.
The startup gives AI agents something simple but surprisingly missing: a phone number. Through one API, developers can let agents handle calls and messages without stitching together telecom providers, routing systems, voice infrastructure, and compliance layers manually.
The idea came from watching AI evolve faster online than in the real world. Meet Previously built AI Agent infrastructure on WhatsApp for 280+ million businesses at Meta. Manav earlier helped grow Vogue’s app from around 100,000 users to more than one million users. Both saw the same gap forming as AI agents became more capable.
Most communication still happens through calls and texts. Businesses run sales, support, recruiting, logistics, and verification flows through phone systems built decades before AI existed. AgentPhone believes agents cannot become truly useful if they remain trapped inside chat windows and browser tabs.
That thinking shaped the product around simplicity. Developers can provision US and Canadian phone numbers instantly through one API call. Calls and messages arrive through the same webhook, while AgentPhone handles transcription and text-to-speech in the background. Instead of building telecom infrastructure from scratch, developers can focus directly on the agent experience.
The startup also built native MCP support, allowing tools like Claude Code, Cursor, OpenClaw, and Windsurf to work with AgentPhone directly. Teams can integrate through Python, Node.js, REST APIs, or MCP servers depending on how they already build agents.
Early use cases already feel closer to real businesses than AI demos. Some teams are building AI recruiters that screen candidates over calls and messages. Others are creating AI employees that manage workflows end-to-end through phone interactions. AgentPhone says teams connected to Google’s Agent Development Kit, Replit, LangChain, Alchemy, and Y Combinator are already building with the product.
AgentPhone joined Y Combinator as part of the Spring 2026 batch with a two-person team based in San Francisco. While much of the AI industry focuses on making models smarter, Meet and Manav are betting something more basic may matter first: giving agents a trusted identity people can actually reach.
I write about startups, AI, technology, and the people building the future. I enjoy breaking down product launches, startup ideas, and tech shifts into simple insights that are easy to understand. My focus is on what’s changing, why it matters, and how technology is shaping the next generation of companies. Through my writing, I aim to help readers stay informed and think deeper about innovation.