Cycling through all six. Tap any point to stop.

Measured on six things

Let your AI agents work in the apps and accounts already on your iPhone — from your Mac, with no Xcode or developer setup.

Security

Header probe scores 1/12: no enforced HTTPS, no HSTS, no CSP, no frame protection, no DMARC. Local-only architecture and Keychain passcode storage are real strengths, but the public site is wide open.

Screenshots

1 / 5

Zeptap

About this app

Zeptap is a Mac app that lets AI agents like Claude Code, Codex and Cursor see, tap, swipe and type on your real iPhone — no Xcode, no Developer Mode, nothing installed on the phone. Hand your agent the phone work you keep picking up your iPhone for. Prepare Instagram Stories and TikTok drafts, reply to customer messages, move an approved video and caption from WhatsApp into Instagram, or have your coding agent test the app you're building on a real device. Chat in Zeptap with your own Claude Code or Codex subscription, or connect any MCP agent. Watch it work from your Mac, ask it to pause before posting, sending or paying, and run up to five iPhones at once. Monthly or lifetime licenses; AI access is separate.

Key features

1Drive a real iPhone from Claude Code, Codex, Cursor, or any MCP agent — no Xcode, no Developer Mode, nothing installed on the phone.
2Local-only architecture: MCP server on loopback (127.0.0.1:47801), passcode in macOS Keychain, no Zeptap cloud or account.
3Bring your own AI: chat runs on your existing Claude or ChatGPT plan, or an OpenRouter key — no credits, no token markup.

Target customer profile

Drawn from the product itself, not from a survey.

Demographic

AI agent power users and developers

Pain points

Simulators and device farms can't touch real accounts, 2FA, or App Store apps

Primary needs

Drive real iPhone apps from Claude Code, Codex, or Cursor with zero dev setup

Demographic

Social media and growth marketers

Pain points

Preparing posts, drafts and replies inside mobile-only apps takes hours of manual phone work

Primary needs

Have an AI agent prepare drafts and replies in the apps on their iPhone, pausing before anything is posted or sent

Demographic

QA and mobile app testers

Pain points

Device farms are pricey, simulators miss real-device behavior and real apps

Primary needs

Automate real-device testing via MCP without Xcode, signing, or Developer Mode

Scout Review

Written by AI from measured evidence, scored out of 100.

Zeptap is a Mac app that lets AI agents drive a real iPhone — tap, swipe, type, read the screen — with no Xcode, no Developer Mode, and nothing installed on the phone. The pitch is sharp and the docs are unusually good for a product this young. The site is polished, the free first hour is a smart trial, and the local-only architecture is a real privacy story. The gaps are concrete: mobile page speed is sluggish, the public site is missing every standard security header, and there's no accessibility statement. For agent power users and mobile QA teams tired of simulators, the wedge is real. For everyone else, it's a niche tool with a niche price.

Measure by measure

Usability

Deep, agent-readable docs with install, pairing, tool reference, playbooks and troubleshooting. Free first hour unlocked, no account needed, license key by email. Onboarding path is unusually clear.

Design

Polished, cohesive landing page with a strong three-iPhone-in-Mac-window motif and consistent blue-accent palette. Hero CTAs compete for attention and device-screen text is too small and low-contrast to read.

Speed

Desktop Lighthouse 98/100, but mobile 68/100 with LCP 6.8s and CLS 0.155 — both past the 'good' thresholds. Fast on a Mac, sluggish on a phone, which is ironic for a product about phones.

Security

Accessibility

Lighthouse accessibility 96/100 with one failing audit: insufficient color contrast. No WCAG statement or a11y page found. Markdown docs and llms.txt help machine and assistive reading.

Growth

Sharp wedge: real iPhone control for agents with no Xcode, no Developer Mode, no device install. Clear segment (agent power users, mobile QA, social marketers) and concrete pricing. Category is early but real.

Improvement areas

Public site security headers: enforce HTTPS, add HSTS, CSP, X-Content-Type-Options, Referrer-Policy, and DMARC (probe scores 1/12).
Mobile performance: LCP 6.8s and CLS 0.155 are past Google's 'good' thresholds — compress hero assets and reserve layout space.
Accessibility: fix the color-contrast failure flagged by Lighthouse and add a WCAG conformance statement to the site.

Design & Usability

Zeptap's site is well-crafted for its stage. The three-iPhone-in-a-Mac-window hero is a strong, repeatable motif, the palette is consistent, and the feature sections scan cleanly. The weak spots are real but small: the two hero CTAs fight for attention, and the device-screen text in the mockup is too small and low-contrast to actually read. Usability is where Zeptap punches above its weight. The docs are genuinely deep — install, pairing, agent connection, tool reference, playbooks, troubleshooting — and they ship markdown and llms.txt endpoints so agents can read them directly. The free first hour with everything unlocked, no account, and license-key-by-email activation is a clean onboarding path. For a product this new, the documentation quality is the standout signal.

Speed & Security

Speed is split: desktop Lighthouse 98/100, mobile 68/100 with LCP 6.8s and CLS 0.155. The mobile numbers are past Google's 'good' thresholds, which matters for a product whose audience is likely browsing on a phone. Security is the weakest dimension. The HTTP probe scores 1/12 — no enforced HTTPS, no HSTS, no CSP, no frame protection, no DMARC. The product's own architecture is more careful: the MCP server binds to loopback only, the passcode lives in the macOS Keychain, and nothing is installed on the iPhone. But the public site serving the download is missing every standard hardening header, which is a real, fixable gap.

Accessibility & Growth

Accessibility scores 96/100 on Lighthouse with a single failing audit: insufficient color contrast. No WCAG statement or dedicated a11y page was found, but the markdown docs and llms.txt endpoints are a genuine assistive-reading win. Growth is a judgement on the idea. Zeptap names a real problem — simulators and device farms can't touch real accounts, 2FA, or App Store apps — and offers a specific wedge: real iPhone control with no Xcode, no Developer Mode, no install on the phone. The target segments (agent power users, mobile QA, social marketers) are named and plausible. The alternatives list shows the field: BrowserStack for device farms, TapKit for agent iPhone control, Buffer for social scheduling. Zeptap's differentiator is the zero-setup real-device angle, which is concrete and defensible. The category is early, and the domain is brand new, so the ceiling depends on execution.

Conclusion

Zeptap is early, focused, and honest about what it is: a bridge between your AI agent and your actual iPhone. The docs are the best part — deep, agent-readable, and clearly written by someone who actually uses the thing. The pricing is straightforward: $25/mo, $99 lifetime, $199 for up to five iPhones. The free first hour is a fair trial. If you're already paying for Claude Code or Codex and you've ever wished your agent could just tap the damn button on your phone, this is the tool. If you're not, the $99 lifetime is a reasonable bet on a category that's still figuring itself out.

Suggestions for improvement

Enforce HTTPS and add HSTS, CSP, X-Content-Type-Options, Referrer-Policy, and DMARC to the public site — the probe scores 1/12.
Fix mobile LCP (6.8s) and CLS (0.155) — compress hero images and reserve layout space for the device mockups.
Fix the color-contrast failure flagged by Lighthouse and increase the size/contrast of text inside the hero device mockups.
Add an accessibility statement and a WCAG conformance target to the site; the docs are already markdown-friendly, so this is low effort.

How we score

Alternatives comparison

Named competitors, point by point. Nobody paid to appear here or to be left out.

	Zeptap	BrowserStack	TapKit	Buffer
Real iPhone control for AI agents	Yes — core feature; drives a real iPhone via MCP with no Xcode or Developer Mode	No — cloud device farm for manual and automated testing, not agent-driven real-phone control	Yes — gives AI agents control of an iPhone; closest direct competitor	No — social media scheduling and publishing, not device control
Requires Developer Mode or Xcode	No — explicitly avoids Developer Mode, Xcode, code signing, and Apple Developer fee	No — cloud-hosted devices, no local dev setup required	Not verified from fetched content	No — SaaS scheduling tool, no device setup
Pricing model	$25/mo, $99 lifetime (1 iPhone), $199 lifetime plus (up to 5 iPhones); free first hour	Subscription tiers based on parallel sessions and minutes; not verified in this review	Not verified from fetched content	Free tier plus paid plans per social channel; not verified in this review
Bring your own AI model	Yes — uses your Claude Code, Codex, or OpenRouter key; no credits or token markup	No — testing platform, not an AI agent host	Not verified from fetched content	No — scheduling tool, not an AI agent host
Local-only / no cloud account	Yes — no Zeptap account, no cloud; MCP server listens on loopback only	No — cloud platform; devices are hosted remotely	Not verified from fetched content	No — cloud SaaS; requires an account