// the find
firerpa/lamda
Android Full-Stack Device Control Platform: WebRTC/H.264 remote desktop, UI/OCR/image-matching automation, one-click MITM, built-in Frida, proxy/VPN/frp/P2P networking, MCP/Agent, 160+ APIs, designed for multi-device clusters and engineered deployments.
FIRERPA is an all-in-one Android device control platform that bundles remote desktop, UI automation, traffic interception, Frida, proxy/VPN, and an MCP server into a single on-device service with a Python client library. It targets security researchers, QA engineers running device farms, and anyone who has previously stitched together Appium + mitmproxy + frida-server + adb into a fragile shell-script stack. Six years of active development and Android 6–16 support give it real staying power.
The virtual display feature is genuinely useful — running automation on a background display without touching the main screen is something Appium cannot do at all. One-click MITM with automatic CA install, Android-version-aware proxy setup, and clean teardown removes the most tedious part of mobile traffic analysis. Bundling frida-server with stealth patches maintained by the project means you don't chase Frida version mismatches every release cycle. The protobuf-over-gRPC client/server design (visible in the rpc/ directory) is a sane foundation for fleet control, and the mDNS device discovery plus frp/P2P networking support make multi-device setups workable without a custom VPN.
The Python client library is a thin wrapper around gRPC stubs with minimal documentation beyond the README — the actual API surface of 160+ methods is largely undiscovered unless you read the proto files or run the examples, which makes onboarding slow. The on-device server binary is closed source; you're trusting a prebuilt ARM binary with root access, no reproducible build, and the repo ships root.crt and root.key in plaintext under tools/ which is a bad look for a security-focused tool. The MCP integration appears bolted on rather than designed — the extension examples are thin and the agent mode depends on any OpenAI-compatible endpoint being available, so it falls apart in air-gapped or corporate environments. License is proprietary for commercial use despite the GitHub presence, and the pricing/subscription model for the networking service (hub, P2P) is opaque.