Sales prep
Research a company or contact on the web, then save a meeting prep brief before the call.
LiveAutoPilotAI: browser, computer and Android automation agents Agents that execute.
Not just answer.
Automation agents for the browser, the computer, and Android, carrying multi-step work through to a finished result.
Files, reports, and outputs you can use as they are. Self-hosted, on your own machine.
Screenshots
1 / 10
What it does
Agents across the browser, your computer, and Android. Chain them, then schedule them.
Browser Agent
Navigates sites, searches, fills forms, and extracts data, then returns it as a structured report.
System Agent
Organizes files and reads PDFs, spreadsheets, and Office documents on your own machine.
Android Agent
Drives connected Android devices over ADB: app navigation, screenshots, and on-device work.
Flow automation
Chain agents into multi-step flows and set them to run daily, or on a schedule you define.
14 ready templates
Shopping, news briefings, sales preparation, competitive research, QA. A running start.
Full transparency
Every run is saved as a task with step-by-step actions, reasoning, timing, screenshots where relevant, and output files you can download.
Self-hosted
Runs on your own machine with Node.js, MongoDB, and Chrome. Bring your own OpenRouter API key for the models.
Who it's for
For anyone carrying work that repeats.
Operators & analysts
Hand repetitive browser and file-based research to an agent.
Founders
Turn recurring work into scheduled flows that run without a browser tab open.
Developers & builders
Automate QA runs and release checklists, stress-test with repeated flows, and use the direct agent consoles.
How it works
How AutoPilotAI works
Running, in 4 steps.
- 01
Describe the workflow
Say what needs doing, in plain language.
- 02
The agent plans and runs
It works out the steps, opens what it needs, and completes the task on your browser, computer, or Android device.
- 03
Schedule and repeat
Run once, later, on a schedule by day, time, and timezone, or several times in a row. The scheduler runs inside the backend, so keep it running.
- 04
Audit and adjust
Review each run's step-by-step log, screenshots, and output files, then refine the instructions and run it again.
Use cases
What AutoPilotAI is used for
Where it earns its place.
Cross-app reporting
Chain browser, system, and Android steps in one flow, feed each step's output into the next, and save one report on a weekly schedule.
Finance and admin
Pull receipt files from a folder, categorize spend, and produce a monthly expense rollup.
QA testing
Repeat a browser path several times in a row to catch flaky UI, or run an Android smoke checklist on a device or emulator and log pass or fail.
Daily briefings
Morning news, market snapshots, and risk updates saved as markdown reports on a weekday schedule.
Comparison
AutoPilotAI vs. alternatives
How AutoPilotAI compares.
| vs. | AutoPilotAI | Alternative |
|---|---|---|
| RPA suites | Plans each step at run time from a screenshot and the page's DOM | Replay steps recorded in advance |
| Integration platforms | Drives the interface, no API required | Limited to integrations that exist |
| Hosted computer-use agents | Self-hosted, on your infrastructure, Android included | Run on the vendor's servers |
FAQ
Frequently asked questions
Common questions about AutoPilotAI.
- What is AutoPilotAI?
- AutoPilotAI is an automation platform whose agents carry out multi-step work in the browser, on your computer through files and shell commands, and on Android: clicking, typing, navigating, and finishing the task the way a person would.
- How is AutoPilotAI different from RPA suites?
- AutoPilotAI does not replay a recorded script. Its agents decide each step at run time from the current state: a screenshot plus the page's DOM in the browser, and a screenshot plus the tappable elements on Android. The browser agent is hybrid: it treats the screenshot as ground truth, acts through DOM selectors, JavaScript, and typing, and switches strategy after an action fails twice. You describe the goal in plain language instead of recording each click.
- Can AutoPilotAI control an Android phone?
- Yes. AutoPilotAI's Android agent controls a physical device or an emulator (Android Studio AVD, Genymotion, or anything else adb can see) from the computer running the backend. It opens and installs apps, reads SMS, uses deep links, navigates the UI, and takes screenshots. It needs Android SDK platform-tools (adb) on that computer and USB debugging enabled on the device. The AutoPilotAI dashboard and backend run on the computer, not the phone.
- Is AutoPilotAI self-hosted?
- Yes. AutoPilotAI runs on your own computer: a Next.js dashboard and a Node.js backend, with tasks, flows, and history in your MongoDB database (local or Atlas) and output files on disk. Model calls go to OpenRouter with your own API key, so the prompts and screenshots each agent sends to the model pass through OpenRouter.
- What can I automate with AutoPilotAI?
- AutoPilotAI handles price comparison and price watches, daily news and market briefings, company research and meeting prep, competitor sweeps, receipt and expense rollups, support queue triage, browser and Android QA runs, content repurposing, trip planning, and Downloads-folder cleanup. Results are saved as files: markdown reports, CSVs, spreadsheets, charts, and presentations.
- Do I need to code to use AutoPilotAI?
- No. You describe a task in plain language in AutoPilotAI's Chat, and its orchestrator routes it to the browser, system, or Android agent. Flows are built from plain-language steps too. Installing it does take a terminal: Node.js 18 or later, MongoDB, and Google Chrome, with one-command setup scripts for macOS, Linux, and Windows.
- What do I need to run AutoPilotAI?
- AutoPilotAI needs Node.js 18 or later, MongoDB (local or Atlas), Google Chrome on the same computer as the backend, and an OpenRouter API key. The Android agent also needs adb and a device with USB debugging, or an emulator. Google sign-in is optional; email and password login works without it.
- Can AutoPilotAI use different models for different agents?
- Yes. AutoPilotAI lets you set a separate OpenRouter model for chat routing, the browser agent, the Android agent, and the system agent, from presets or a custom model ID.
- Can AutoPilotAI notify me when a flow finishes?
- Yes. AutoPilotAI can send email alerts through Brevo and signed webhooks (HMAC-SHA256 in an X-Webhook-Signature header) for nine flow and task lifecycle events, each toggled in Settings.
- Does AutoPilotAI keep me logged in to websites?
- It can. AutoPilotAI's browser agent can use a dedicated Chrome profile so logins persist across runs, or a temporary session for a clean slate.
- Is AutoPilotAI safe to run on my computer?
- AutoPilotAI runs shell commands, reads files, uses a real Chrome profile, and controls a phone, so run it only on machines and accounts you trust. Its system agent blocks a list of high-risk commands such as rm -rf / and mkfs.
- Where can I buy AutoPilotAI?
- AutoPilotAI is sold on Wixzel Store. The download includes the backend and frontend source code, one-command setup scripts for macOS, Linux, and Windows, and an installation guide.