Technology

How Airtap works.

Three layers. AI in the cloud, hands on every device, drop-in for any agent that reads SKILLS.md.

Architecture

Brain · Hands · Devices

The brain decides. AutoPilot does. The action lands on a real or cloud phone.

Brain · Hands · Devices The brain decides. AutoPilot does. The action lands on a real or cloud phone. OPTIONAL Your AI Agent CLAUDE · CODEX · OPENCLAW Drops in via SKILLS.md VIA SKILLS.md THE BRAIN Airtap AI Cloud BRAIN · MEMORY · ROUTINES Decides what should happen, when THE HANDS AutoPilot EXECUTES · TAP · SCROLL · TYPE Cloud Phone Dedicated phone in the cloud Always on · Apps stay logged in NO SETUP · INSTANT Physical Devices Your real device, your apps AutoPilot app installed ANDROID · iOS · TV · PC
Three layers

What each layer does

1 · Brain
Airtap AI Cloud

Memory, routines, and decisions. Persistent across sessions, always reachable. Or — drop in your own agent.

2 · Hands
AutoPilot

Turns intent into actual taps. Operates real apps — tap, scroll, type, navigate, exactly like a person would.

3 · Device
Cloud Phone or yours

AutoPilot runs on a dedicated cloud phone (always on) or on your physical device — your apps, your accounts. Or both.

Two device paths

Cloud Phone or your phone

Cloud
Cloud Phone

Every Airtap user gets a dedicated phone in the cloud. Apps stay signed in. Tasks run 24×7 — without tying up your device, without depending on your battery or network. Best for monitoring, scheduled jobs, and overnight automation.

Physical
Your phone + AutoPilot

Install AutoPilot on Android, iOS, TV, or PC. Airtap can then drive the apps you already use, on the device you already carry. Best for actions tied to a specific account, location, or hardware.

Use either. Or both. Same brain, same routines.

Agent integration

Drop in Claude, Codex, or OpenClaw

One SKILLS.md file teaches your agent how to operate a phone. Your agent stops just answering and starts acting.

1
Get the skill file

Download Airtap's SKILLS.md — a single file that defines how an agent operates a phone.

2
Load it into your agent

Drop the file into Claude, Codex, OpenClaw, or any SKILLS.md-compatible runtime.

3
Assign a phone

Spin up a cloud phone or connect a physical device. Your agent is ready to act.

What this enables

Once an agent has a phone

Messaging
Send and reply

Across WhatsApp, Telegram, iMessage, Slack — the agent reads, the agent writes.

Social
Post and monitor

TikTok, Instagram, X — publish, track engagement, respond to comments.

Booking
Reserve and buy

OpenTable, ClassPass, Amazon — claim the table, the class, the deal at the moment it opens.

Watching
Monitor feeds

Track price drops, claim status, waiver clears — and act when the trigger hits.

Workflows
Multi-step tasks

Chain apps end-to-end — gather, decide, execute, confirm — without dropping context.

Multi-agent
A network of phones

Each agent gets its own phone, identity, and persistent app sessions. Coordinate across them.

See it in action →

Watch real demos of agents driving real apps.

Brain in the cloud.
Hands on every phone.

Try Airtap