Mano-P
Mano-P: Open-source GUI-VLA agent for edge devices. #1 on OSWorld (specialized, 58.2%). Runs locally on Apple M4 Mac mini/MacBook — no data leaves your device.Mano-P 是一个开源 GUI-VLA 项目，支持在 Mac mini/MacBook 上或通过算力棒本地运行推理，实现纯视觉驱动的跨平台 GUI 自动化操作。数据完全本地处理，支持复杂多步骤任务规划与执行。
Mano-P is an open-source GUI-VLA (Vision-Language-Action) agent skill that enables AI assistants to perform desktop automation through pure visual understanding. It runs locally on Apple Silicon devices (M4+ with 32GB RAM) or via compute stick, processing all screenshots and tasks on-device without cloud API calls. The skill integrates with Claude Code and OpenClaw through ClawHub, allowing agents to autonomously execute multi-step GUI operations across browsers, desktop apps, and native macOS interfaces. Developers can use it either as a CLI tool (mano-cua) for direct invocation or as an agent skill that LLM agents call autonomously during reasoning.
Key Features
Use Cases
- 01Enable Claude Code agents to autonomously interact with desktop apps and web UIs during development workflows
- 02Automate data extraction and integration across multiple applications without API access
- 03Execute enterprise business process automation with dozens of GUI steps while keeping data on-device
- 04Drive end-to-end browser testing in CI/CD pipelines (e.g., Mano-AFK's E2E test stage)
- 05Perform complex office automation tasks (spreadsheets, documents, email) through natural language commands
- 06Build privacy-preserving GUI automation workflows for sensitive data environments
Related Skills
View moresuperpowers
An agentic skills framework & software development methodology that works.
skills
Skills for Real Engineers. Straight from my .agents directory.
skills
Public repository for Agent Skills
ponytail
Makes your AI agent think like the laziest senior dev in the room. The best code is the code you never wrote.
Mano-P — FAQ
What is Mano-P and how does it work?+
Mano-P is a GUI automation agent skill that uses vision-language models to understand and control desktop interfaces through screenshots. It can run locally on Apple M4+ devices (32GB+ RAM) or connect to cloud inference, translating natural language tasks into GUI actions like clicks, typing, and navigation.
How do I install Mano-P as a CLI tool?+
Install via Homebrew: brew tap Mininglamp-AI/tap && brew install mano-cua. For local inference, run mano-cua check, mano-cua install-sdk, and mano-cua install-model. Grant Screen Recording and Accessibility permissions in System Preferences.
How do I add Mano-P to Claude Code or OpenClaw?+
Install the skill from ClawHub: clawhub install mano-cua, or manually download from clawhub.ai/hanningwang/mano-cua and copy to the commands directory. Restart Claude Code or start a new session to enable the skill.
What are the hardware requirements for local mode?+
Requires Mac mini or MacBook with Apple M4 chip (or newer) and 32GB+ RAM. Alternatively, any Mac can use the Mano-P compute stick via USB 4.0+ connection. Cloud mode works on any system with internet access.
Is Mano-P free and does it require API keys?+
Mano-P is open-source (Apache 2.0 license) and free for commercial use. Local mode requires no API keys or internet. Cloud mode sends screenshots to mano.mininglamp.com (no API key needed) but processes data externally.
Which AI clients can use Mano-P?+
Mano-P works with Claude Code and OpenClaw via ClawHub skill installation. The CLI tool (mano-cua) can be invoked directly from terminal or embedded in shell scripts and CI/CD pipelines.
How do I install Mano-P?+
Open the source repository on GitHub and follow its README. Mano-P is a skill — MCP Agents Market links you directly to the official repo.
Is Mano-P free?+
Mano-P is an open-source project hosted on GitHub. Check the repository for its license and any usage requirements.