A desk lit by a warm lamp in the evening, an open laptop and a mug beside a stack of papers

TapKit: Let Claude and Codex Control Your Own iPhone

TapKit is a Mac app that gives AI agents your iPhone. You plug a real iPhone into your Mac, and TapKit lets Claude, Codex or your own code see its screen and tap, swipe and type across any app on it, through a hosted MCP server or a REST API. Nothing is installed on the phone, and no jailbreak or developer mode is needed. Because it’s your own phone, the agent works with your apps, accounts, iMessage and iCloud as they already are. TapKit costs $49 a month per phone and is made by Jootsing Research Inc.

The quick answer: TapKit takes the bring-your-own-device route to phone use for agents. Instead of renting a phone in a data center, you turn an iPhone you already own into something an agent can operate, from your desk or a closet. That makes it quick to start and strong for personal tasks and iOS testing, but it means keeping a Mac and an unlocked, plugged-in iPhone running. A disclosure: we make Carly, an AI assistant that makes and answers calls and texts, which is a different job from operating a phone’s apps. We’ve kept to TapKit’s site, docs and policies.

TapKit at a glance

Checked against TapKit’s site, documentation, acceptable use policy and subprocessor list on November 12, 2026.

TapKit
What it isA Mac app that turns your own iPhones into a tool AI agents can control
How agents connectA hosted MCP server, a REST API with a Python SDK, and a built-in agent in the Mac app
Works withClaude Code, the Claude app, Codex and any MCP client with OAuth
HardwareYour own iPhone, a USB data cable and a Mac on macOS 14.6 or later
PhonesUp to seven iPhones per Mac, plus phones shared by your team
Price$49 a month per phone
CompanyJootsing Research Inc., United States

What TapKit does

TapKit’s homepage pitches it simply: “Give your agent your iPhone.” Its promise is that anything you can do on your phone, your agent can do too, and it lists what that includes: downloading new apps, texting with iMessage, handling two-factor verification, using your iCloud and paying for things. “Every app becomes a tool your agent can use. Your agent sees the same screen you do.”

TapKit groups its uses into three areas:

  • Personal assistant. Booking appointments, ordering dinner and handling everyday tasks in the apps you already use, including helping with iMessage conversations and checking out with your saved details while you confirm purchases yourself with Apple Pay.
  • Account management. Authorized account setup, verification, support, booking and tasks that cross several apps.
  • App development. Letting an agent explore other apps or test yours with real iOS features, across iPhones shared by your team.

How it works

TapKit’s docs describe three parts: your iPhone, the Mac app and the TapKit cloud.

The Mac hosts the iPhone. The iPhone stays plugged into a Mac running TapKit. The Mac app is the only part of TapKit that talks to the phone. It uses the Mac’s camera permission to see the iPhone’s screen over the cable, never the Mac’s webcam, and pairs with the iPhone over Bluetooth, with AssistiveTouch turned on, to send taps, swipes, typing and key presses.

Agents reach the Mac through the cloud. While you’re signed in, the Mac app keeps a connection open to TapKit’s cloud. Claude, Codex or your own code sends commands through MCP or the REST API, and the cloud forwards each one to the right Mac and sends the result back. That means the agent can be anywhere, and “the Mac and phone can be in a closet, a data center, or on your desk.”

Every agent works in a loop. It takes a screenshot, decides what to do, acts, then takes another screenshot to check the result. TapKit’s MCP server returns JPEG screenshots scaled to at most 1,344 pixels on the long edge, and the agent taps by pixel coordinates in that image.

The docs explain why TapKit is built this way: “No jailbreak, no sideloading, no developer mode. Nothing is installed on the iPhone.” And because it works through the screen a person sees, it “works with any app” rather than relying on app-specific integrations.

Connecting your agent

TapKit has several ways in:

  • MCP. A hosted server at mcp.tapkit.ai for Claude Code, the Claude app (Chat and Cowork), Codex and any MCP client that supports OAuth with dynamic client registration. You approve each client on a consent screen.
  • REST API. For custom integrations, with a Python SDK and API keys from the web app.
  • Skills. Instructions that teach agents how to use TapKit and get around iOS.
  • Built-in agent. The Mac app has its own agent, which works with the phone directly and calls a hosted language model.
  • Live view. You can watch and control a phone from the Mac app or the web app, including phones connected to a teammate’s Mac.

Setup

Setup takes a few minutes in the Mac app. You install it, sign in with Google, allow camera and Bluetooth access, plug in and unlock your iPhone, tap Trust, and pair the iPhone with the Mac over Bluetooth. TapKit turns on AssistiveTouch for you. Each extra iPhone you plug in goes through the same steps and takes one slot on your subscription.

Two practical requirements follow from using a real phone:

  • The phone has to stay unlocked. “TapKit can’t unlock a phone, so the phone has to stay unlocked.” The docs suggest setting Auto-Lock to Never so it doesn’t lock between tasks.
  • The phone has to stay plugged in. To cut TapKit off, you unplug the phone, quit the Mac app or forget the Mac in Bluetooth settings.

Pricing

TapKit’s pricing is per phone: $49 a month for each iPhone, so one phone is $49 a month and five are $245 a month. Every plan includes the full product, you can cancel anytime, and larger fleets with setup and custom integrations are priced on a call. Since you supply the phones and the Mac, there’s no hardware rental on top.

Security and privacy

TapKit’s security page is clear about what it can and can’t do.

  • It can’t get past your locks. “TapKit can’t get past the passcode, Face ID, or anything else that would stop a person holding the phone.”
  • Screenshots go to your AI provider. With MCP and the API, commands and screenshots travel through TapKit’s servers, “and TapKit stores them.” Your agent’s provider, such as Anthropic or OpenAI, receives every screenshot your agent requests.
  • The screen leaves your Mac only when needed. The Mac app sends the screen elsewhere only when an agent takes a screenshot or someone opens a live stream, which is handled by Agora.
  • Credentials are protected. The Mac app keeps its session in the macOS Keychain, MCP access tokens expire after an hour, and TapKit stores only a hash of each API key.
  • Shared access. Everyone in your organization can use its phones.

TapKit is also honest about the main risk. Its skills tell agents to check what they type, stop before submitting unless you’ve asked them to, and not type passwords you haven’t given them. “These rules reduce risk but don’t remove it, because an agent with your phone can do anything you can. Use a phone and accounts you’re comfortable handing to an agent.”

Its subprocessors include Anthropic and OpenAI for AI models, Agora for streaming, and Stripe, Supabase and Vercel.

Rules on what you can do with it

TapKit’s acceptable use policy limits it to “devices that you own or that you are expressly authorized to access, control, and automate,” and prohibits covert surveillance of another person’s device. Among other things, it bans creating fake accounts, generating artificial engagement or manipulating reviews and ratings, spam, and circumventing “an account restriction, suspension, ban, or other enforcement action imposed by a Third-Party Application.” It also prohibits misrepresenting automated activity as human “where a platform or applicable law requires automation to be disclosed.” You’re responsible for anything your agents do under your account.

The company

TapKit is made by Jootsing Research Inc., a US company. Its Mac app has shipped frequently since version 1.0 in January 2026, with more than 70 releases through October 2026. TapKit says it’s exploring hosted phone options for teams that don’t want to manage their own hardware.

How TapKit compares

TapKit is a bring-your-own-phone option among services giving agents phones; the open-source Phone Harness also drives your own iPhone from a Mac, through Apple’s iPhone Mirroring instead of a cable. Glasshand and PhoneUse rent you physical iPhones in a data center, and Phonebox and Droidrun offer cloud Android phones. TapKit keeps everything on your own hardware, so your agent uses the exact phone, apps and accounts you do, without renting a device. The tradeoff is that you manage the Mac, the cable and an unlocked phone yourself.

If the job is phone calls and texts rather than working through apps, Carly does phone and SMS out of the box: it makes and answers calls and texts from a real business number and handles email and calendars across all your accounts, from inside ChatGPT or Claude or by text.

Who TapKit suits

TapKit is a good fit if:

  • You want Claude or Codex to work on your own iPhone, with your apps and accounts already signed in.
  • You already have a spare iPhone and a Mac that can stay on.
  • You build iOS apps and want an agent to test them on a real phone.
  • You want several iPhones, up to seven per Mac, each doing its own task.

Look elsewhere if:

  • You don’t want to manage hardware or keep a phone unlocked and plugged in.
  • You need Android, or many phones on demand.
  • You’d rather not give an agent a phone with your personal accounts on it.
  • You need a hosted device today; TapKit says hosted phones are still being explored.

FAQ

What is TapKit?

TapKit is a Mac app from Jootsing Research Inc. that turns your own iPhone into a tool AI agents can control. Agents see the screen and tap, swipe and type through a hosted MCP server or a REST API.

How much does TapKit cost?

$49 a month per iPhone. Five phones cost $245 a month, and larger fleets are priced on a call.

Does TapKit work with Claude?

Yes. TapKit’s hosted MCP server works with Claude Code and the Claude app, as well as Codex and other MCP clients that support OAuth.

Do I need to jailbreak my iPhone?

No. TapKit installs nothing on the iPhone and doesn’t need a jailbreak or developer mode. It sees the screen over the cable and sends input over Bluetooth with AssistiveTouch.

Can TapKit unlock my iPhone?

No. TapKit can’t get past the passcode or Face ID, so the phone has to stay unlocked while an agent uses it.

Do I need my own iPhone?

Yes. TapKit is bring-your-own-device. You connect your own iPhone to your own Mac.

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