Photon Raises $4.5M to Route AI Agents Through iMessage and WhatsApp, Bypassing App Stores Entirely

Messaging platforms are becoming the new operating systems for artificial intelligence agents, and a startup called Photon is building the infrastructure to make it happen.

The company closed a $4.5 million seed funding round co-led by Gradient and A*, with participation from Vercel, HongShan, Z Fellows, Llama Ventures, Karman, and undisclosed angel investors. The capital follows a four-month stretch where Photon reported a tenfold increase in revenue and amassed more than 40,000 developer sign-ups for its platform, which routes AI agent interactions through iMessage, WhatsApp, Telegram, SMS, RCS, email, and voice channels.

For developers building conversational AI tools, distribution has historically been the most expensive hurdle. Publishing an app requires navigating Apple’s App Store or Google Play, paying up to 30% commission on digital goods, and spending heavily on user acquisition to convince consumers to download yet another standalone application. By routing agents directly through messaging apps that billions of people already open daily, Photon eliminates the discovery bottleneck entirely.

This approach mirrors the original promise of chatbots during the mid-2010s, when platforms like Facebook Messenger attempted to replace native apps with conversational interfaces. That wave largely failed because the underlying natural language processing was too rigid to handle complex requests. Today’s large language models have fundamentally changed that equation, allowing agents to manage multi-step workflows—from booking travel to processing insurance claims—within a standard text thread.

Photon provides a unified API, a command-line interface, an extensible channel framework, and an observability suite to support these deployments. The managed platform guarantees 99.95% uptime and carries SOC 2 Type II and HIPAA compliance certifications. Those security credentials are critical for enterprises handling protected health information or financial data over messaging protocols, a requirement that separates commercial deployments from hobbyist projects.

The company’s origins trace back to an open-source tool created by CEO Daniel Tian and CTO Ryan Zhu while they were students. Tian built a prototype that could automatically reply to his friends via iMessage. After publishing the code on GitHub, thousands of developers adopted it. Both founders subsequently left their academic programs—Tian dropped out of the University of Pennsylvania’s Management and Technology program, and Zhu left high school before becoming a visiting student at MIT’s Media Lab—to build the company full-time.

That open-source foundation remains central to Photon’s footprint. According to Tian, 98% of current usage still runs on the free, open-source version of the software. The commercial hosted tier, introduced in April, restricts free access to 10 users, pushing larger deployments into three paid subscription tiers. The company reports a churn rate below 3% and a fivefold increase in messaging volume over the past month alone.

Commercial adoption spans several sectors. Fintech customers include business banking platform Rho and finance assistant Fliptexts. Consumer applications using the infrastructure range from dating platform Ditto to social introductions app Boardy and AI email client Slashy. Corgi Insurance represents the company’s push into regulated industries.

Photon has also embedded itself into the broader AI engineering stack. Integration partners include Nous Research, which uses Photon as the default iMessage layer for its Hermes agent. Tencent relies on the technology for QClaw and NanoClaw. Framework-level integrations extend across LangChain, Mastra, Convex, Render, Railway, and Telnyx.

To underscore its thesis that traditional mobile applications face obsolescence, Photon staged a literal funeral for app icons inside a San Francisco church on September 17. The event, which featured a physical coffin for app logos alongside technical panels from engineers at Stripe, OpenAI, and Vercel, doubled as a developer day.

While the immediate market involves humans interacting with AI over text messages, the architecture points toward a different scale of operations. As autonomous software becomes capable of executing specialized tasks, individual agents will need to communicate with one another to complete complex workflows. A personal assistant coordinating a trip, for example, would need to negotiate directly with separate flight-booking and hotel-reservation agents without requiring human intervention at each step. This agent-to-agent communication layer represents the next phase of infrastructure development for companies positioning themselves at the intersection of messaging and machine intelligence.