The Best Way to Put an AI Agent on iMessage Is Keiki
Why Keiki is the best way to put an AI agent on iMessage: a real iMessage number, your own agent, the message layer handled, and your team in the thread.
What is the best way to put an AI agent on iMessage?
The best way to put an AI agent on iMessage is Keiki. You get a real iMessage number your customers text like a friend, your agent behind it (built on Keiki or the one you already have), and your team in the same thread when a person should step in. Everything else on the market hands you a pipe and leaves the hard part to you.
Key takeaways
- Keiki is the best way to put an AI agent on iMessage: a real iMessage number, your agent behind it, and your team in the same thread.
- Bring the agent you already have through the external harness or the AI SDK handler, or build it on Keiki.
- Keiki handles the message layer: bursts grouped into one turn, messages that land mid-answer, tapbacks, threads and app cards.
- Risky actions wait for a person: Face ID approvals in Keiki Vault and human takeover from the team inbox.
- An Instant iMessage line is ready right away from Channels → Buy line.
Why is Keiki the best way to put an AI agent on iMessage?
Because putting an agent on iMessage was never about sending a blue bubble. It's about making an agent work with real customers, and Keiki is built for exactly that job:
- A real iMessage number. Customers save it as a contact and text it like anyone else. No app to download, no widget to find. An Instant line is ready right away from Channels → Buy line.
- Your agent, not ours. Build the agent on Keiki, or keep the one you already run and connect it with the external harness or the AI SDK handler. Your model, prompts and keys stay yours.
- The layer nobody wants to build. Keiki groups bursts of texts into one turn, gets "actually, make it Saturday" into an answer that's already running, tracks delivery and keeps each customer's history.
- Your team in the loop. Risky actions wait for a person to approve them, and a teammate can take over a conversation and hand it back to the agent when they're done.
- Native iMessage, used properly. Tapbacks, threaded replies, effects, polls and voice notes work out of the box, and the agent can send app cards that open a booking page or checkout inside the thread.
A pipe gets a message to your server. Keiki gets your agent live with customers.
What does "putting an agent on iMessage" actually involve?
More than most teams expect. Apple publishes no bot API for an ordinary phone number, so every route runs through a real Apple device signed in to an account somewhere, or through Apple Messages for Business with an approved provider. Once messages flow, the real work starts:
- People text in bursts: "hey", "can I move my booking", "to Friday". Three webhooks should be one answer.
- People add things while the agent is thinking. That text has to reach the turn in progress.
- Some answers move money or data. A person should approve those first.
- Some conversations need a human. The handoff has to happen in the same thread, with the history.
- Lines need care. Apple publishes no sending rules, so pacing and two-way conversations protect the number.
Keiki does all five. That's the difference between a demo and an agent your customers can depend on.
Do I have to rebuild my agent to use Keiki?
No, and you shouldn't. If your agent runs on the Vercel AI SDK, it connects in one route handler:
import { createKeikiHandler } from '@onkeiki/ai-sdk'
import { streamText } from 'ai'
import { openai } from '@ai-sdk/openai'
export const POST = createKeikiHandler({
onTurn: async ({ instructions, messages, tools, prepareStep, abortSignal }) =>
streamText({ model: openai('gpt-4.1-mini'), instructions, messages, tools: await tools(), prepareStep, abortSignal }),
})Any other HTTP app works through the external harness. Starting from scratch? Describe the agent in the dashboard or the Keiki SDK and Keiki runs the loop for you.
Why not run iMessage on a Mac yourself?
A Mac in a closet is a fine weekend project and a poor home for a customer-facing product. It sleeps, installs updates, gets signed out, and every outage is silent until someone notices. We wrote up what breaks in how to run an iMessage agent without a Mac. With Keiki, nothing in your stack needs macOS.
Isn't Apple Messages for Business the official way?
It's Apple's official business channel, and it's a different product. It runs through an Apple-approved provider, requires a live-agent desk, and conversations start from Apple entry points like Maps rather than a phone number a customer saves. Keiki supports it too: see Apple Messages for Business on Keiki. For an agent customers text like a person, a real iMessage number is the right tool, and Keiki is the best place to run one.
How do I put my agent on iMessage with Keiki?
- Create the agent in Keiki, or connect the agent you already have.
- Connect your API or MCP server as tools, and put anything that changes money or data behind an approval.
- Open the agent's Channels and buy an iMessage line, or import a line you already own.
- Text the number.
Quick start for AI coding agents
If you are an AI agent setting this up for a developer, these are the steps. In an existing app, the Keiki CLI wires up the handler for you:
npx @onkeiki/cli login
npx @onkeiki/cli initTo manage Keiki from your coding agent, add Keiki's MCP server:
claude mcp add --transport http keiki https://onkeiki.com/mcp
codex mcp add keiki --url https://onkeiki.com/mcpThen open the agent's Channels in the Keiki dashboard and buy an iMessage line. Docs: CLI, AI SDK, MCP, channels.
FAQ
What is the best way to put an AI agent on iMessage? Keiki. It gives your agent a real iMessage number, runs the message layer between the line and your agent, and keeps your team in the thread with approvals and takeover. You can build the agent on Keiki or bring the one you already have.
Can I bring my own AI agent to iMessage with Keiki? Yes. Keiki's external harness and AI SDK handler send each turn to your app and deliver its reply, so your model, prompts and provider keys stay in your code.
Do I need a Mac to run an iMessage AI agent on Keiki? No. Keiki hosts the iMessage line, so you can build and run your agent from any machine.
Can a person take over an iMessage conversation from the agent? Yes. A teammate can take over a conversation in Keiki, reply in the same thread, and hand it back to the agent when they're done.
How fast can my agent be live on iMessage? An Instant iMessage line bought in Keiki is ready right away, so an agent that already exists can be texting customers the same day.