clawd-conduit
Plugin / AgentWire all 15 Claude Code hook events into the Clawd on Desk desktop pet — and turn permission requests into a blocking desktop card.
- hooks
- desktop
- notifications
- claude-code
A Claude Code plugin that runs 20 AI testers on 20 phones in parallel on free GitHub runners — one Claude Code agent per persona on iOS Simulators and Android emulators, using your app like a new user and writing bug reports with screenshots
You fork the template into your own repo (gh repo create --template), run setup.sh once (key pair, secrets and Claude token), then trigger agent-testers.yml: one workflow builds the app, then spins up one emulator or simulator per persona, each driven by a Claude Code agent that gets three tools — see the screen (with tap coordinates), gestures and typing. Personas set the device: model, OS, font size, dark mode, language and location. Reports and recordings are encrypted with your key before they land in logs, and analytics hosts are blocked inside the phones so tests never touch your real stats. Agents read the screen from screenshots and the accessibility tree, so any UI toolkit works — labelled widgets (Flutter Semantics, Android contentDescription, accessibilityLabel) buy them more precise tap targets. Frameworks are detected automatically: Flutter, React Native/Expo, native Kotlin/Java and Swift/Objective-C, Kotlin Multiplatform, Capacitor/Ionic, .NET MAUI (beta), or your own build command via the android_build/ios_build inputs. The README's own demo run: the demo app carries 11 planted bugs, 4 testers found 9 of them in 2-7 minutes each (about 21 minutes for the whole run, builds included) plus unplanted real issues, at $0.87 of model usage for the 4.
Use when you ship a mobile app and want a cheap, repeatable smoke pass before every release: parallel persona-based exploratory testing on real simulator/emulator builds, bug reports with screenshots in your repo, and no extra test-infra bill — at the cost of Claude usage from the agents' turns.
Template flow (the README's quick start): gh repo create my-testers --public --template pierridotite/mobile-agent-testers --clone cd my-testers && ./setup.sh # key pair + secrets + Claude token (Windows: .\setup.ps1) gh workflow run agent-testers.yml # demo: 2 Android + 2 iOS testers, ~20 min bash scripts/collect.sh <run id> # decrypts reports into runs/<id>/SUMMARY.md Or use it as a Claude Code plugin: /plugin marketplace add pierridotite/mobile-agent-testers /plugin install mobile-agent-testers@mobile-agent-testers Run your own app: gh workflow run agent-testers.yml -f source_repo=you/app -f app_id=com.you.app -f android_personas=anna,malik -f ios_personas=clara,jonas
The free-lunch math is documented, not implied: GitHub's hosted runners are free for public repos with about 20 concurrent jobs and only 5 on macOS, 20 agents can burn through a Claude subscription's 5-hour window in about 15 minutes (the README says run batches of 6-10), and hosted runners are meant for the repo's own project — for a closed-source app, run the workflow in your private repo or on your own Mac (a private repo needs a read-only SOURCE_REPO_TOKEN secret to pull your app). Results are encrypted precisely because public-repo logs are public. Model spend scales with testers: the demo's $0.87 covered 4 agents finding 9 of 11 planted bugs. iOS runs need arm64 Xcode builds on macOS runners (5 concurrent). MIT. 29 stars as of writing (2026-11-18).
Repository
pierridotite/mobile-agent-testersCharted
Wire all 15 Claude Code hook events into the Clawd on Desk desktop pet — and turn permission requests into a blocking desktop card.
An unofficial Rust 1:1 reimplementation of the Claude Code terminal UI, ported file-for-file onto an iocraft TUI framework.
Hook-level enforcement for the natural-language rules in AGENTS.md and CLAUDE.md — every edit is judged and violations are repaired in the same turn.