A woman on a phone call at a kitchen counter in the evening, with a notebook and a mug of coffee beside her

Photon: Open-Source iMessage Infrastructure for Agents

Photon is a San Francisco startup that connects AI agents to the messaging apps people already use, starting with iMessage. Its core is an open-source TypeScript framework called Spectrum: you write your agent once, and Spectrum delivers it over iMessage, WhatsApp, Telegram, SMS and RCS, or voice calls. You can run it free on your own Mac or pay Photon for managed phone lines.

The quick answer: Photon is the developer’s option among the iMessage providers. It’s open source, you can start free, and one codebase reaches iMessage, WhatsApp and Telegram. Its shared plans give each of your users a different pooled number, and only the dedicated plans get group chats and limited outbound. It’s also the youngest company in the group, with a hosted platform that launched in April 2026. A note before you read on: we make Carly, which sells AI calling and texting for businesses, so we overlap a little with what Photon’s customers build. Everything below is sourced and linked.

Photon at a glance

Checked against Photon’s site, docs, terms and GitHub on November 12, 2026.

Photon
What it isAn open-source agent messaging framework (Spectrum) plus hosted iMessage lines
ChannelsiMessage, SMS and RCS fallback, WhatsApp Business, Telegram, voice over SIP
FoundersDaniel Tian (CEO) and Ryan Zhu (CTO)
Funding$4.5 million seed, October 2026, co-led by Gradient and A*
PlansSelf-hosted free, Free (10 users), Pro $25/month (100 users), Business $250 per line per month, Enterprise custom
Open sourcespectrum-ts (Apache 2.0, about 2,200 GitHub stars), imessage-kit (MIT, about 1,500 stars)
ComplianceSOC 2 Type II and HIPAA support, per Photon
Best forDevelopers and startups putting an agent into iMessage, especially early on

Who builds on Photon

Photon’s showcased customers are mostly consumer agents that live in a text thread. Ditto is an iMessage matchmaker for college students that Photon says has connected more than 42,000 users, and Flip is an AI financial assistant you text to move money and track a portfolio. Photon also lists Fae, an “AI best friend” in iMessage with 85% week-four retention according to Photon, the startup bank Rho, and Corgi Insurance and Boardy. The AI Insider reported that Photon also powers Nous Research’s Hermes agent.

How a student iMessage bot became Photon

Tian and Zhu started Photon after struggling to get anyone to discover the apps they built as students, according to The AI Insider. Their answer was to skip the app entirely and meet people in iMessage. They built an open-source iMessage bot, and thousands of developers picked it up.

That code still lives on GitHub. Photon’s imessage-kit, a TypeScript SDK for iMessage on macOS, was published in October 2025 and has about 1,500 stars. Spectrum, the multi-channel framework that followed in December 2025, has about 2,200 stars and is under active development. The hosted platform came in April 2026.

In October 2026, Photon announced a $4.5 million seed round co-led by Gradient and A*, with Vercel, HSG, Z Fellows, Llama Ventures, Karman and angel investors joining. It reported more than 40,000 developer sign-ups, revenue up tenfold in four months, churn under 3%, and the open-source bot still accounting for 98% of usage. To mark its thesis that agents will replace apps, the company held a tongue-in-cheek “app funeral” in San Francisco in September.

How Photon works

Spectrum is the layer your agent code talks to. You register providers, and one message handler receives text, attachments, reactions and other events from all of them. Sending works the same way: you can let Spectrum pick the best channel for a recipient, or route explicitly, for example iMessage first with Telegram as the fallback.

The providers Photon documents today:

  • iMessage (cloud): managed lines run by Photon, connected over gRPC.
  • iMessage (local): a separate macOS-only provider that reads the Messages database on your own Mac.
  • WhatsApp Business: through Meta’s official WhatsApp Business Cloud API.
  • Telegram: the Telegram Bot API.
  • Voice: place and receive calls over SIP, using a Spectrum iMessage line.
  • Terminal: a local chat interface for building and testing.

On iMessage, the feature list includes polls and chat backgrounds from iOS 26, message effects, attachments, live location sharing, group chats, contact cards and what Photon calls generative iMessage mini apps, small interactive experiences inside the conversation. Several of those are limited to the Business plan. Webhooks are signed with HMAC. Photon 2.0, in beta, is a newer unified API that adds email and other channels.

Three ways to get an iMessage number

This is where Photon differs most from its rivals.

Self-hosted on your Mac. Run the open-source code on a Mac and it uses your own iCloud account as the agent’s iMessage number. It’s free and good for building locally, but your personal account becomes the agent.

Managed shared (Free and Pro). Photon keeps a large pool of numbers and, in its words, “assigns each of your users a fresh one they have never received a message from before.” So two of your users may see your agent text from two different numbers. Shared mode can’t create group chats.

Dedicated (Business). Numbers allocated to your project alone. Every user texts the same number, group chats work, and with auto-scale on, Photon adds another line when one approaches capacity.

Photon pricing

Photon’s pricing page charges by end users on the shared plans and by line on Business.

PlanPriceNumberLimits and features
Open sourceFreeYour own (via your Mac)Self-hosted, limited group support
Free$0Managed sharedUp to 10 users, polls and chat backgrounds, community support
Pro$25/monthManaged sharedUp to 100 users, fast-track support tickets
Business$250 per line per monthManaged dedicatedUnlimited users with auto-scale, group chats, location sharing, mini apps, contact sharing, cold outreach up to 50 new contacts a day, WhatsApp and phone calls
EnterpriseCustomDedicated, and you own the numbersContractual SLA, lowest throttling, dedicated Slack and Discord support

Every iMessage plan includes RCS and SMS fallback and “unlimited daily messages,” but unlimited messaging and unlimited users both require auto-scale. On Business, Photon says auto-scale means “your account automatically adds credits when needed,” and the pricing page doesn’t list what those credits cost. Ask before you turn it on.

Paid plans renew monthly, and you can cancel or delete your account at any time, per Photon’s terms.

The limits that keep a line healthy

Photon’s deliverability guide is unusually specific. It starts from the point that iMessage is encrypted, so Apple “can’t read message content: they filter on behavior.” Two quotas are enforced:

  • 5,000 outbound messages per server per day. Photon calls this a hard limit and says going past it raises the risk of Apple banning the line.
  • 50 new conversations per line per day. A new conversation is the first message to someone the line has never messaged.

It also recommends 400 to 500 users per line for moderate conversation, 200 to 400 for heavy use, and no new users on a server past 70 to 80% utilization.

The design advice will sound familiar if you’ve read Linq’s docs. Let users text first, since Apple shows a “Report Junk” banner on messages from unknown numbers. Ship sms: links with a pre-filled message so the user is the one who hits send. Share a contact card early. And aim for at least three messages from each user. Photon says that after a user sends three messages, Apple treats the conversation as trusted. Its list of what gets lines flagged starts with burst sending (100 or more messages from one line in a short window), broadcasting without conversation, and repeatedly messaging people who don’t reply.

Photon also claims “the lowest ban rate we have observed” among iMessage providers, while adding that “no provider can remove Apple or carrier limits.” That’s a company claim we can’t check.

Security, data and Apple

Photon’s homepage says it is SOC 2 Type II compliant and offers “HIPAA-compliant support,” and The AI Insider reported the same for its hosted platform. If you need either, ask for the report and a BAA.

Two things are worth raising before you send sensitive messages. In October 2026, a developer opened a GitHub issue asking Photon to document “how is relayed message content stored, accessed and protected?” It was still open when we checked. And Photon’s terms let it use usage and aggregated data to “develop, train, fine-tune, evaluate, and improve Photon’s machine-learning models.” Its privacy policy says personal data is deleted or anonymized within 30 days of account deletion, with exceptions for legal and business needs.

Like every service in this category, Photon’s managed lines depend on Apple tolerating them. Apple’s official business channel is Apple Messages for Business, and Photon wasn’t on Apple’s approved provider list when we checked. Photon’s terms say its messaging relies on third-party providers it doesn’t control, subject to their limits and policies.

Photon compared with the other iMessage APIs

PhotonLinqBlooio
Main audienceDevelopers and early-stage startupsAI agent and consumer app buildersSmall businesses, agencies and AI agents
Number optionsYour own Mac, a shared pool, or dedicated linesDedicated lines with managed load balancingShared pool, reply-only or dedicated lines
Free optionSelf-host, or 10 users hostedHobby, 20 contacts20-message trial
Reaching new peopleBusiness: up to 50 a day per lineNot for cold outreachNot allowed
Open sourceYesNoNo
Other channelsWhatsApp, Telegram, voiceVoiceApple Messages for Business, RCS business, SMS and voice

Other managed iMessage APIs include Chert, which adds FaceTime audio and video calls, messages.dev and ZeroSettle. If you want email and phone alongside iMessage under one agent identity, see Inkbox.

Who Photon suits

Photon is a good fit if:

  • You’re a developer or early-stage startup putting an agent into iMessage and want to start free.
  • You want one codebase that also reaches WhatsApp and Telegram, which matters for users outside the US.
  • You like being able to read and self-host the code.

Look elsewhere if:

  • You need every user to see the same number. That takes a dedicated line, since shared plans assign numbers per user.
  • You need clear limits on auto-scale before you launch. Photon doesn’t publish what the credits it adds cost.
  • You need a long track record. Photon’s hosted service is months old.

And if you’re not building an agent but want one working for your business today, answering the phone and texting customers back, that’s Carly out of the box. Add it to ChatGPT or Claude, get a business number, and its agents take calls, text back missed callers and reply on WhatsApp. It uses standard SMS rather than iMessage. Our AI text messaging guide shows what that looks like day to day.

FAQ

Is Photon free?

Yes, in two ways. The open-source framework is free to self-host on your own Mac, and the hosted Free plan covers up to 10 users on a shared number. Pro is $25 a month for up to 100 users.

What is Photon Spectrum?

Spectrum is Photon’s open-source TypeScript framework (spectrum-ts) for connecting one AI agent to iMessage, WhatsApp, Telegram, SMS, RCS and voice through a single message handler.

Can Photon send cold outreach?

Only on the Business and Enterprise plans, and only up to 50 new contacts per line per day. Photon’s own guidance says inbound-first conversations are far safer for line health.

Does Photon support group chats?

Yes, on dedicated lines (Business plan). The shared-pool plans can’t create group chats.

Who founded Photon?

Daniel Tian (CEO) and Ryan Zhu (CTO). The company is based in San Francisco and raised a $4.5 million seed round in October 2026.

Does Photon work with WhatsApp?

Yes. Spectrum includes a WhatsApp provider built on Meta’s official WhatsApp Business Cloud API, and WhatsApp is included on the Business plan.

Ready to automate your busywork?

Carly schedules, researches, and briefs you—so you can focus on what matters.

See what people say

"Before Carly, I relied on a Calendly link, but the whole process felt impersonal and not very professional. Carly changed that by handling all the back-and-forth, so I'm no longer stuck in endless email threads trying to line up schedules.

Now Carly reaches out to candidates, shares my real-time availability, lets them pick a slot, then sends a Zoom link and drops it straight into my calendar. She sends reminders to both of us before each call, which has significantly reduced no-shows and last-minute confusion.

On top of scheduling, Carly acts like a full executive assistant, sending me my schedule the night before so I can prepare for each call. It reminds me of the old x.ai assistant, but Carly is noticeably smarter, faster, and better suited to my healthcare recruitment business."

Gus Ibrahim, Founder & Director, IHR