Automates native macOS apps through DeepChat's Computer Use plugin by driving accessibility APIs and fallback pixel coordinates. You get a standardized loop: resolve the app by bundle ID, launch it, snapshot the window state, act through MCP tools like click or type_text, then verify the result. When accessibility trees are sparse (common in Electron and media apps), it falls back to screenshots and coordinate-based clicks or tries enabling debugging ports for DOM access. The permission model is explicit: you'll need to grant Accessibility and Screen Recording to the helper bundle. Honest take: this bridges the gap between Claude's reasoning and real GUI automation without writing AppleScript or wrestling with brittle selectors, though sparse trees mean you're sometimes clicking pixels and hoping.

GitHub stars
6K
last updated
Jun 11, 2026

Install

$ npx -y skills add thinkinaixyz/deepchat --skill cua-driver --agent claude-code

Installs into .claude/skills of the current project.

Files

cua-driver

Orchestrates macOS app automation via cua-driver. Whenever a user asks to drive a native macOS app, follow the loop in this skill rather than calling tools ad-hoc — the snapshot-before-action invariant is not optional and silently breaks if you skip it.

DeepChat MCP mode

When DeepChat auto-pins this skill because the Computer Use MCP server is enabled, use the available Computer Use tools in the current tool list directly. Treat examples such as click({...}) or get_window_state({...}) as calls to the matching MCP tool with the same JSON payload. Shell CLI examples are diagnostic references for shell-based agents.

App-name resolution belongs here. When matching a user request to an app, compare the request with each list_apps name and bundle_id. Consider the user's language, system language, English product names, English brand names, romanized or pinyin variants, and common abbreviations. Treat bundle_id as the strongest identity signal. When the requested app name is ambiguous, localized, translated, abbreviated, or written in another script, call list_apps, resolve the most credible bundle id, then call launch_app with that bundle id.

Sponsors

Details

Repository
github.com/thinkinaixyz/deepchat
Categories
AI & Agent Building, Automation & Workflows
License
Apache-2.0
First seen
Jun 11, 2026

More from thinkinaixyz/deepchat

All 19
1thinkinaixyz avatarfeishu-toolsthinkinaixyz/deepchatUse the Feishu/Lark plugin MCP tools for Feishu documents, spreadsheets, knowledge content, and other matching workspace operations.0skills.sh installs
2ThinkInAIXYZ avatarmemory-managementThinkInAIXYZ/deepchatGuide the agent to recall, remember, and route durable learning into Memory, Skills, Scheduled Tasks, or Tape.0skills.sh installs
3UKGovernmentBEIS avatarcode-quality-fix-allUKGovernmentBEIS/inspect_evalsFix code quality issues identified in a code quality review stored in agent_artefacts/code_quality/<topic>/. Systematically addresses issues found by the code-q0skills.sh installs
4utopai-research avatarvideo-composeutopai-research/pai-proGenerates and prompts video clips on the filmmaking canvas. Use when the user asks to generate, render, animate, continue, restyle, edit, shoot, or compose a vi0skills.sh installs
5wbern avatartddwbern/claude-instructionsRemind agent about TDD approach and continue conversation0skills.sh installs
6zhixiangluo avatarlinkedin-engagementzhixiangluo/10xproductivityAutomated LinkedIn engagement workflow. The agent finds a relevant post on your chosen topic, drafts a comment with a genuine insight, gets your approval in cha0skills.sh installs