eggbrid2/mobileClaw — Open Android AI agent runtime for phone control, app automation, VLM screen reading, skill routing,…

Description and use

**Summary**
eggbrid2/mobileClaw is an open‑source Android runtime that lets developers build AI agents capable of controlling the phone, automating apps, reading screens with vision‑language models, routing skills, hosting mini‑apps, and chaining Mihomo VPN workflows. It transforms isolated prompts and tools into repeatable, orchestrated agent pipelines, making multi‑agent coordination and tool‑use pipelines easy to standardize and reuse.

**Value**

• **Unified agent workflow engine** – turns ad‑hoc prompts into reproducible, version‑controlled pipelines, reducing engineering overhead for mobile AI automation.
• **End‑to‑end capabilities** – combines UI interaction, VLM screen parsing, skill routing, and VPN management under a single Kotlin SDK, eliminating the need to stitch together disparate libraries.
• **Extensible mini‑app framework** – lets teams plug in custom “mini‑apps” (e.g., data collectors, chat interfaces) that agents can invoke on‑device, fostering rapid prototyping and reuse.

**Practical adoption path**

1. **Proof‑of‑concept** – clone the repo, run the provided README example on a test device, and verify basic screen‑reading and app‑automation functions.
2. **Pilot integration** – replace a manual automation script with a mobileClaw skill, using the built‑in skill‑routing API to connect VLM‑based screen parsing to the new skill.
3. **Scale to multi‑agent flows** – define a workflow YAML (or Kotlin DSL) that chains multiple skills (e.g., VPN setup → app login → data extraction) and store the definition in version control for CI/CD testing.
4. **Production hardening** – add persistent agent memory (SQLite or encrypted storage), enforce permission policies, and integrate with your existing monitoring/observability stack.

**Production readiness**

• **Activity & community** – 400 ★, 16 forks, recent commits (last update 2026‑06‑26), and a Kotlin codebase with 17 relevant topics indicate a healthy, actively maintained project.
• **Ecosystem fit** – aligns with orchestration, automation, and mobile AI use cases; the SDK is lightweight enough for on‑device deployment yet extensible for backend orchestration.
• **Risk considerations** – no obvious licensing or metadata red flags, but a final security audit (dependency scanning, permission review) and confirmation of active maintainers are recommended before a full‑scale rollout.

Overall, mobileClaw is mature enough for a serious pilot and, with the outlined incremental integration steps, can be promoted to production for robust mobile AI agent workflows.

Problem

eggbrid2/mobileClaw helps turn isolated prompts and tools into repeatable agent workflows.

Outcome

High for an OSS candidate: recent activity, adoption, and ecosystem signals are strong enough for a serious pilot.

Workflow

• coordinate multi-agent workflows
• add tool-use pipelines
• standardize agent memory

Evidence

Open Android AI agent runtime for phone control, app automation, VLM screen reading, skill routing, mini apps, and Mihomo VPN workflows.

Source