Photon Bets Messaging Beats App Stores With $4.5M Seed
A San Francisco startup helping developers build agents that run on iMessage and WhatsApp has closed seed funding after 40,000 sign-ups and 10× revenue growth in four months.
The Distribution Wedge
Every time Daniel Tian and Ryan Zhu finished building a new consumer app at hackathons, they ran into the same wall: nobody could find it. The App Store's search ranking, the cold-start problem of social graphs, the friction of convincing someone to install yet another icon on their home screen all added up to a distribution puzzle that no amount of clever product design could solve.
So the two co-founders flipped the problem. Instead of asking users to come to a new app, they built software that lives inside the messaging apps users already open dozens of times a day. Photon, the infrastructure company that emerged from that realisation, announced it has closed a $4.5 million seed round co-led by Gradient and A*, with participation from Vercel, HongShan, Z Fellows, Llama Ventures, Karman, and a group of angel investors.
The funding arrives after Photon signed up more than 40,000 developers and grew revenue tenfold over four months, according to the company. Tian, who left the University of Pennsylvania's M&T programme to pursue the venture full-time, and Zhu, who departed high school and later became a visiting student at MIT's Media Lab, are betting that the next generation of software will not be distributed through app stores at all.
What Photon Actually Does
Photon provides a stack of developer tools for building and operating agents that work across iMessage, WhatsApp, Telegram, SMS, RCS, email, and voice channels. The core offering includes a unified API that abstracts away the quirks of each messaging platform, an extensible channel framework, a command-line interface, and an observability suite for monitoring agent behaviour in production.
The open-source version of Photon still accounts for 98 per cent of usage. When Tian built a bot that would reply to his friends over iMessage as him and published the code on GitHub, thousands of developers forked it. That organic adoption convinced the pair to commit. In April this year, they launched a managed, hosted platform sold via subscription tiers.
The hosted service offers 99.95 per cent uptime, a meaningful improvement over running the open-source code on a local machine for hobby projects. It also carries SOC 2 Type II and HIPAA compliance, opening the door to healthcare agent use cases that would otherwise be off-limits for a developer toolkit without formal certification.
Any developer using Photon with more than ten end users must move off the free tier onto one of three paid plans. The company declined to share revenue figures but reported less than 3 per cent monthly churn and a fivefold increase in messaging volume over the past month alone.
Who Is Building on It
Photon's customer base spans fintech, consumer AI, and vertical SaaS. Corgi Insurance, social introductions platform Boardy, Gen Z dating app Ditto, business banking service Rho, finance AI assistant Fliptexts, and AI email client Slashy all run agents on Photon infrastructure.
On the technology partner side, Vercel's Eve and ChatSDK projects integrate Photon, as does Nous Research's Hermes agent, which uses Photon as its default iMessage layer. Tencent's QClaw and NanoClaw agents also sit on top of Photon. The startup has secured framework and infrastructure integrations across LangChain, Mastra, Convex, Render, Railway, and Telnyx.
Across both hosted and open-source deployments, Tian estimates the technology now reaches millions of end users. That reach is partly why Vercel, a previous integration partner, joined the cap table in this round.
The Agent-to-Agent Bet
While Photon's current product focuses on connecting agents to humans, Tian sees a larger opportunity in facilitating agent-to-agent communication. In his view, the next phase of the market will involve agents discovering and coordinating with other specialised agents to complete multi-step tasks.
Consider a travel planning scenario: a user's personal AI assistant might pull in a flight-booking agent, a hotel-booking agent, and a local transport agent in sequence, orchestrating the handoffs without requiring the user to switch contexts or interfaces. Photon's messaging abstraction layer could serve as the connective tissue for that kind of agent-to-agent-to-person workflow.
The company describes this emerging pattern as A-to-A-to-P: agent-to-agent-to-person. If that architecture takes hold, the distribution advantage Photon offers today, getting agents in front of users without requiring an app install, extends into a coordination advantage tomorrow, routing tasks between specialised agents that may be operated by different developers or companies.
Why This Matters Now
At Opentechwire, we've tracked a steady drumbeat of investment into agentic infrastructure over the past eighteen months, but Photon's approach is distinct in two respects. First, it solves a distribution problem rather than a capability problem. Most agent startups are building better reasoning, better tool use, or better memory; Photon is building a way for those agents to reach users without competing for App Store rank or ad spend.
Second, the timing aligns with a shift in user behaviour that has been underway since well before the current AI wave. Messaging apps have become super-apps in much of Asia, and even in markets like the United States, where app unbundling dominated the last decade, users are increasingly reluctant to install new software. Meta's recent launch of Muse, which climbed to number one on app stores despite being an AI assistant accessible through existing Meta properties, suggests that even when users do download a new app, they prefer it to integrate with communication tools they already use.
Photon is not claiming that mobile apps will disappear overnight. Tian himself acknowledges the transition will take time. But the company is positioning itself at the intersection of two trends, the rise of agentic software and the plateau of app install growth, that together create a wedge for a new distribution model.
The question is whether developers building the next generation of consumer and enterprise software will choose to distribute through messaging channels at sufficient scale to make Photon's infrastructure a standard layer. With 40,000 developers signed up and integration partnerships across major frameworks, the early signals suggest that bet is not as far-fetched as it might have seemed two years ago.
What Comes Next
Photon's immediate roadmap centres on expanding channel support, improving observability tooling, and scaling the hosted platform to handle the messaging volume growth the company is seeing. Longer term, the agent-to-agent coordination layer represents a more ambitious technical and go-to-market challenge.
If agents do begin to compose and coordinate with one another to complete complex tasks, the infrastructure that brokers those interactions will need to handle authentication, state management, and billing across organisational boundaries. That is a harder problem than providing a unified API for sending messages, and it is one that will likely require Photon to build deeper into the agentic stack.
For now, the company is focused on proving that developers want to build on messaging platforms and that users prefer interacting with agents inside apps they already use. The $4.5 million seed round gives Photon runway to test both hypotheses at scale. Whether the app era is truly ending or simply evolving, the startup has positioned itself to benefit either way.
