longarm MCP & REST automation

longarm turns an Android phone or tablet into a local automation endpoint for developers, QA testers, and AI agents. Run the built-in HTTP API or streamable HTTP MCP server on your device, control a real Android screen from another computer on the same network, or just chat with the in-app AI agent and let it drive the phone itself.

Use longarm to automate app flows, test on physical Android devices, collect screenshots, operate a phone from a desktop tool, or connect an MCP-compatible agent to a real mobile screen instead of relying on emulator-only workflows.

ANDROID AUTOMATION OVER HTTP
- Start a local HTTP API server on your Android phone or tablet
- Send tap, swipe, long-press, pinch, rotate, and two-finger swipe commands to exact screen coordinates
- Launch any installed app by package name, or fire a custom Android intent (URL, deep link, or activity)
- Capture PNG screenshots, and query screen width, height, and density for reliable scripts
- Show a floating overlay with optional live coordinates

BATCH TASKS: SAVE AND REPLAY MULTI-STEP FLOWS
Chain taps, swipes, screenshots, app launches, and intents into a single named batch task instead of firing one-off commands. Add delays, repeat steps, and use loop blocks (count, range, foreach, or while) with variables. Build and run tasks from the Batch tab, or create, update, and trigger them remotely over the same REST API and MCP server — so an agent can script an entire flow in one call.

MCP SERVER FOR AI AGENTS
longarm includes a streamable HTTP MCP server so compatible AI agents and developer tools can call Android control tools directly. Agents can inspect screen information, tap, long-press, open installed apps, launch Android intents, save and run batch tasks, and, with Plus, use advanced gestures such as swipes, pinches, two-finger swipes, and rotation.

IN-APP AI AGENT: CHAT TO CONTROL YOUR PHONE (BYOK)
Describe a goal in plain language in the Agent tab and longarm's built-in AI agent inspects the screen, then taps, swipes, types, opens apps, and runs batch tasks to get it done. Bring your own API key for Anthropic Claude, Google Gemini, or an OpenAI-compatible endpoint, or use the free on-device model — no key, no network call. The agent calls the same tap, swipe, screenshot, app-launch, and batch-task functions exposed over REST and MCP. Plus adds voice input/output and advanced gestures.

REST API DOCS READY FOR SCRIPTS AND AGENTS
longarm includes over 20 documented endpoints, including app/intent launching and full batch-task CRUD. The Docs tab shows live parameters and lets you copy the full API as Markdown, ready to paste into a script, test harness, or AI agent prompt.

DETAILED LIVE LOGS FOR REST AND MCP
The History tab shows incoming REST requests and MCP tool calls with methods, status codes, filters, and expandable params and results — visible even while the server is offline, without exposing stored bearer tokens in plain text.

BATCH RUN HISTORY AND SCREENSHOT RESULTS
Every batch run is recorded with a run ID, status, progress, messages, and captured screenshots, so you can review finished, failed, or aborted workflows and export screenshot evidence for QA and debugging.

NO ROOT, LOCAL NETWORK CONTROL
longarm runs on your LAN and does not require root. Configure bind addresses and ports, require a bearer token, keep the screen awake, and auto-start the server on launch.

FOR DEVELOPERS, TESTERS, AND AI AGENTS
Use longarm when you need:
- Android automation and remote control from another machine
- Real-device testing without a heavy automation stack
- Scriptable taps, swipes, gestures, and screenshots
- REST API and MCP server access for mobile automation
- An in-app AI agent (BYOK or on-device) that can call every automation function itself
- LAN-only control of a phone or tablet

longarm uses Android Accessibility Services to perform gestures and screenshots only in response to your commands, for controlled automation on devices you manage.