Photon raises $4.5M seed for agent tooling that runs over iMessage and WhatsApp

Photon banks $4.5M for agents that live in your chat apps

Photon, a startup whose tooling lets developers build AI agents that operate over iMessage and WhatsApp, has closed a $4.5 million seed round. According to TechCrunch, the round was co-led by Gradient and A*, with participation from Vercel — itself a Photon technology partner — alongside HongShan, Z Fellows, Llama Ventures, Karman and several angel investors.

The funding follows a period of steep growth. TechCrunch reports that Photon has signed up more than 40,000 developers, says revenue grew tenfold in four months, and is seeing churn below 3%, with messaging volume up fivefold in the past month alone.

A funeral for mobile apps

Photon marked its argument with a publicity stunt: a genuine funeral for mobile apps, held in a San Francisco church on September 17, complete with speeches and a coffin reserved for app icons. The event doubled as a developer day featuring panels from Vercel, Stripe and OpenAI.

Co-founder and CEO Daniel Tian acknowledges the transition will be gradual but argues the direction is fixed: fewer standalone apps, more agents reachable through the messaging services people already open every day. TechCrunch notes that Meta's Muse, currently the top download on app stores, is a live test of that theory. Even if Muse comes to dominate consumer AI, Tian sees a lasting market in giving developers the infrastructure to run their own agents over chat channels.

What the platform provides

Photon's toolkit includes a unified API, an extensible channel framework, a command-line interface and an observability suite. Together, these let teams build and operate agent experiences that work across iMessage, WhatsApp, Telegram, SMS or RCS, email, voice and other channels.

From hackathon side project to hosted platform

The idea grew out of the founders' experience building consumer apps as students, often at hackathons, where they repeatedly hit the same wall: nobody could discover what they built. Tian had written a bot that replied to his friends on his behalf over iMessage; when he and CTO Ryan Zhu open-sourced the technology on GitHub, thousands of developers adopted it. Tian then left UPenn's M&T program, and Zhu left high school, later becoming a visiting student at MIT's Media Lab.

The open-source version still accounts for 98% of Photon's usage. In April, the founders added a managed platform sold through subscription tiers, offering 99.95% uptime — a meaningful step up from running the open-source project on a personal laptop. The hosted product is SOC 2 Type II and HIPAA-compliant, which makes healthcare agents a viable use case. Usage beyond 10 users moves customers off the free tier onto one of three paid plans.

Who is building on it

TechCrunch lists customers including Corgi Insurance, social introductions app Boardy, dating platform Ditto, business banking platform Rho, finance assistant Fliptexts and AI email client Slashy. Integration partners include Vercel's Eve and ChatSDK, and Nous Research, whose Hermes agent uses Photon as its default iMessage layer. Photon also underpins Tencent's QClaw and NanoClaw, and has framework and infrastructure integrations with LangChain, Mastra, Convex, Render, Railway and Telnyx. Across the open-source and hosted versions, Tian estimates the technology reaches millions of end users.

Why it matters

Photon is one of the clearest expressions of the "agents replace apps" thesis: rather than fighting for app-store visibility, developers meet users inside chat threads they already use daily. The round, the developer growth and the partner list suggest the idea is gaining real traction, not just conference-stage rhetoric.

The longer-term bet is larger still. Tian expects agents to start finding other agents — a travel-planning assistant pulling in a flight-booking or hotel-booking agent when needed — which would reposition Photon as an agent-to-agent communication layer rather than a simple bridge between software and people. If that shift plays out, the open question is not just whether apps fade, but which layer ends up routing traffic between agents.