Deploying a Local GUI Agent on Apple Silicon

Description and use

Installing and running the Mano‑P local GUI agent on an Apple Silicon Mac: step‑by‑step Homebrew installation, permission setup, and local task execution without cloud dependencies.

Problem

How to install and run Mano-P, a vision‑driven GUI agent, on an Apple Silicon Mac while keeping all processing local.

Outcome

Mano‑P runs locally on an M5 Pro with ~80 tok/s decode speed and 56% task completion on a 100‑task benchmark, outperforming cloud‑based Qwen3‑VL‑Plus (39%).

Workflow

• brew tap Mininglamp-AI/tap && brew install mano-cua
• mano-cua check
• mano-cua install-sdk
• mano-cua install-model
• Grant Screen Recording and Accessibility permissions in System Settings
• mano-cua run "open Safari, search for today's weather" --local

Evidence

Mano‑P is a vision‑driven GUI agent. It looks at your screen, decides what to click or type, and executes multi‑step tasks autonomously.

Source

The model is Mano‑CUA‑4B‑Thinking‑1.1, quantized to MLX 8‑bit, with a peak memory footprint of 4.3GB.

Source

Hardware requirement: M4 chip or newer with 32GB+ RAM. Alternatively, any Mac with a USB 4.0+ compute stick works.

Source

You'll need to grant Screen Recording and Accessibility permissions in System Settings. That's the entire setup.

Source

On an M5 Pro, decode speed sits around 80 tok/s. In our 100-task real Mac GUI benchmark on a MacBook Pro M5 with 16GB, the Thinking-4B model hit 56% task completion.

Source