Photon Raises $4.5 Million to Put AI Agents Inside WhatsApp and iMessage
AI infrastructure startup Photon has raised $4.5 million in seed funding to pursue a simple but potentially disruptive idea: instead of asking users to download another app, let AI agents reach them through the messaging platforms they already use every day.
The round was co-led by Gradient and A with participation from Vercel, HongShan, Z Fellows, Llama Ventures, Karman and angel investors.
Photon is building infrastructure that allows developers to deploy AI agents across iMessage, WhatsApp, Telegram, SMS and RCS, Slack, Discord and other communication channels from a common framework.
The startup says its platform has attracted tens of thousands of developers, while revenue has increased 10-fold in just four months. Messaging volume also jumped fivefold during the latest month reported by the company.
Those numbers are giving Photon an opportunity to test a broader thesis: the next major interface for AI may not be another chatbot app at all. It could simply be the message thread already sitting on a user’s phone.
Photon Is Betting Against the App Download
Photon’s founders, Daniel Tian and Ryan Zhu, arrived at the idea after repeatedly confronting the same problem while building consumer applications: creating software was easier than persuading people to discover and install it.
Their answer was to remove the download entirely.
An AI service built using Photon can appear directly inside a messaging conversation, allowing users to interact with it much like they would communicate with another person or business.
For developers, that potentially removes several layers of friction.
There is no separate application for users to discover, no unfamiliar interface to learn and, depending on the service, potentially much less onboarding before the first interaction.
Photon’s argument is that messaging applications have already solved one of the hardest problems facing consumer software: getting people to open them every day.
Spectrum Lets Developers Build Once and Reach Multiple Messaging Platforms
At the center of Photon’s technology is Spectrum, its open-source framework and cloud platform for connecting AI agents to messaging interfaces.
Instead of developers separately building and maintaining integrations for WhatsApp, iMessage, Telegram and other channels, Spectrum provides a common infrastructure layer.
Photon combines that framework with a unified API, command-line tools and an observability suite designed to help developers monitor and operate their agents once they reach production.
The infrastructure also handles some of the less visible challenges involved in turning an AI prototype into a functioning messaging product, including message delivery, platform-specific behavior, conversation state and reliability.
That distinction is important.
Photon is not primarily trying to build the AI assistant that consumers ultimately use. It wants to provide the infrastructure through which thousands of other AI assistants reach those consumers.
WhatsApp Gives Photon Access to Billions of Existing Users
The scale of existing messaging platforms explains why that infrastructure could become valuable.
WhatsApp has more than 3 billion users globally, giving developers access to an enormous potential audience without requiring customers to install another piece of software.
Photon is an official Meta Business Technology Provider and connects agents to WhatsApp Business through Meta’s Cloud API.
The platform supports one-to-one conversations, approved messaging templates, media, contextual replies, reactions and click-to-WhatsApp campaigns.
For businesses, that could allow AI agents to operate inside the same communication channel already used for customer service, sales, authentication and other interactions.
The company is applying a similar philosophy to iMessage, where it provides infrastructure allowing agents to send and receive messages without developers having to maintain their own dedicated Mac hardware.
Open Source Still Accounts for 98% of Photon Usage
Photon’s commercial opportunity comes with an important caveat.
The company originally released its technology as an open-source project, and that version still accounts for approximately 98% of usage.
The founders introduced a managed commercial platform in April, offering developers infrastructure that can be used without operating the underlying systems themselves.
Photon says the hosted platform provides 99.95% uptime and has achieved SOC 2 Type II and HIPAA compliance, potentially opening the technology to more sensitive enterprise and healthcare applications.
The company has also introduced paid subscription tiers once projects grow beyond limited free usage.
That means Photon’s next challenge is not simply attracting developers — something its open-source technology has already demonstrated it can do — but converting enough of that developer activity into a sustainable hosted infrastructure business.
More Than 40,000 Developers Have Already Signed Up
Photon says more than 40,000 developers have signed up for its platform, while Gradient, one of the investors leading the new funding round, puts the developer community at more than 50,000.
The difference likely reflects different measurement dates, but both figures indicate rapid early adoption.
Customers and users of Photon’s infrastructure span financial services, insurance and consumer AI.
Companies identified as users include Corgi Insurance, Boardy, Ditto, Rho, Fliptexts and Slashy.
Photon has also established technology integrations across a wider AI development ecosystem.
Its infrastructure is used with tools and frameworks including Vercel, LangChain, Mastra, Convex, Render, Railway and Telnyx, while Nous Research’s Hermes agent uses Photon as its default iMessage layer.
The technology also provides messaging infrastructure for projects including Tencent’s QClaw and NanoClaw.
The Next Step Could Be Agents Talking to Other Agents
Photon’s longer-term ambition extends beyond connecting an AI agent with a human user.
Tian believes messaging infrastructure could eventually become a communication layer between different agents.
A personal AI assistant helping someone organize a trip, for example, might communicate with another agent specializing in flights and another handling hotels before returning the completed options to the user.
That would shift Photon’s role again.
Instead of merely transporting messages between humans and AI, its infrastructure could eventually help route interactions among multiple specialized agents working together on a task.
The company refers to this emerging model as agent-to-agent-to-person communication.
Photon Even Held a Funeral for Mobile Apps
Photon has made its position on the future of traditional apps unusually visible.
On September 17, the company held what it called an “app funeral” at a church in San Francisco, complete with a coffin containing app icons.
The event was partly a publicity stunt but also served as Photon’s developer gathering, with participants from companies including Vercel, Stripe and OpenAI.
Tian acknowledges that traditional applications are not about to disappear overnight.
But the event captured the company’s central argument: as AI agents become capable of performing more tasks directly through conversation, users may have fewer reasons to install and navigate a dedicated application for every individual service.
The Bigger Battle Is Over the Interface for AI
Photon’s $4.5 million funding round is relatively small compared with the enormous sums flowing into AI models and computing infrastructure.
But the company is attacking a different layer of the AI stack.
Large language models determine what an agent can understand and do. Cloud and semiconductor companies provide the computing power. Photon wants to control part of the infrastructure that determines where users actually encounter those agents.
If AI agents increasingly move into WhatsApp, iMessage and other existing communication channels, the competitive battle around artificial intelligence could shift beyond who has the most capable model.
It could also become a battle over who owns the infrastructure connecting those models to billions of everyday conversations.
Photon’s biggest bet, therefore, is not that mobile apps will suddenly disappear. It is that AI could make downloading a separate app unnecessary for a growing number of tasks — and that the humble message thread could become one of the most important interfaces of the agent era.