Lamda

FIRERPA (formerly Lamda) is aimed at a narrow but real audience: people who already own Android hardware and want deep, scriptable access to it without building their own connectivity and instrumentation stack. The browser remote desktop is the entry point — WebRTC video with hardware encoding, multi-touch, audio, clipboard, drag-and-drop upload and click-to-download, plus a web terminal with ADB. That combination removes the usual desktop client and the usual "blind command execution" problem, since you can see the screen while you type into the shell. Above that sit the capabilities that justify the tool for security and QA work: built-in Frida with Hook RPC, persistent scripts and spawn injection; a one-click MITM proxy with per-app capture, certificate setup and upstream proxying; and mobile proxy mode, which exits traffic through the device's own network so a phone on a cellular link reproduces conditions a lab machine never will. The Python API is the automation surface — app operations, file I/O, command execution, KV storage, system properties, device status and service logs — with UI selectors (text, resourceId, description) and OCR/image matching (PaddleOCR, EasyOCR, SIFT, template matching) for screens that expose no usable layout. Cross-network management is handled by STARLINK for pools spread across offices and data centers, with PIGEON dispatching work by device state. The AI story is unusually concrete for this category: built-in MCP server endpoints let Claude, Cursor, or any compatible client observe and act on devices, and PIGEON can hand exception handling to a model when a deterministic selector fails. Release cadence is visible in the changelog: 10.0 (background virtual screen automation, non-root execution, WebRTC mirroring, dark mode, 2GB uploads), 10.2 (Android 17, 16KB page size, Python 3.12, APK/XAPK install API), 10.4 (sqlalchemy, paramiko, fuser/killall/pstree, human-friendly query semantics), 10.6 (openssl and ssh upgrades, bundled mitmproxy library, smaller binaries), and 10.8 (start_activity int64 fix, OpenAI/Anthropic/Ollama support libraries, lower memory use). Where it doesn't fit: it is Android-only, so iOS work needs a different tool. Setup is not trivial — you install the firerpa.apk service, handle Shizuku for non-root devices, and configure autostart whitelists so the service survives reboot; full reach means Magisk or ROM-level deployment. Automation is Python-first with no record-and-replay low-code path, which is a strength for engineers and a wall for everyone else. You supply and maintain the hardware; FIRERPA keeps it reachable. We could not reach the vendor's pricing page during this refresh, so treat the tier figures in our pricing section as last-known values rather than confirmed current ones, and confirm plan details with the vendor before you budget.