Skip to content
SPECIFICTION
Get listed

agent-cu

github.com

Free plan WindowsmacOSLinux
Visit github.com
Spec score 8.1 #3 of 24 in AI Computer Use Agents

What it is

agent-cu is a command-line tool for controlling desktop apps by clicking, typing, and reading interface information. It uses the accessibility tree that screen readers rely on to identify interface elements, then acts on them through references. Commands include snapshots, selectors, clicks, typing, scrolling, dragging, window management, and JSON output. Users can batch commands and reuse YAML workflows. The tool supports Electron apps through CDP, including Slack, Cursor, VS Code, Postman, Discord, and Notion. Its skill can be used with Claude Code, Cursor, Codex, Copilot, OpenCode, Cline, and 40+ other agents. agent-cu runs locally on macOS, Linux, and Windows; precompiled binaries are available for macOS Apple Silicon and Intel, Linux, and Windows. Installation instructions cover npm, pnpm, yarn, and bun. It requires explicit operating-system accessibility permissions to read and control apps. The project is free and MIT licensed. Its repository states that it is no longer maintained and recommends Cua and Cua Driver instead.

Spec sheet

What it does
agent-cu is a CLI for controlling desktop apps by clicking, typing, and reading what is on screen.Read 8 Oct 2026 · agent-computer-use.dev
Intended users
The README describes the tool as intended for AI agents such as Claude, Cursor, and Codex, local models, and humans.Read 8 Oct 2026 · github.com
Security
The security policy says the tool does not transmit data over the network, store credentials or sensitive data, execute remote code, or bypass OS security boundaries.Read 8 Oct 2026 · github.com
Support
The security policy asks users to email maintainers about vulnerabilities and says they will respond within 48 hours.Read 8 Oct 2026 · github.com
License
The project is released under the MIT license.Read 8 Oct 2026 · github.com
Agent integrations
The project says its skill can be used with Claude Code, Cursor, Codex, Copilot, OpenCode, Cline, and 40+ other agents.Read 8 Oct 2026 · github.com
Automation
It supports batched commands and reusable YAML workflows.Read 8 Oct 2026 · github.com
Electron apps
It supports Electron apps including Slack, Cursor, VS Code, Postman, Discord, and Notion through CDP.Read 8 Oct 2026 · github.com
Features
The CLI supports snapshots, selectors, clicks, typing, scrolling, dragging, window management, and JSON output.Read 8 Oct 2026 · github.com
How it works
It reads the accessibility tree used by screen readers to identify interface elements, then acts on them through references.Read 8 Oct 2026 · agent-computer-use.dev
Installation
Precompiled binaries are provided for macOS Apple Silicon and Intel, Linux, and Windows, with npm, pnpm, yarn, and bun installation instructions.Read 8 Oct 2026 · github.com
Permissions
The tool requires explicit OS accessibility permissions to read and control apps.Read 8 Oct 2026 · github.com
Project status
The repository says the project is no longer maintained and recommends Cua and Cua Driver instead.Read 8 Oct 2026 · github.com

Plans and prices

Open source Free MIT licensed · runs locally · zero per-action cost Read 8 Oct 2026 · github.com

Every agent-cu plan and price →

Strong points and limits

Strong points

  • Runs locally on macOS, Linux, and Windows
  • Supports reusable YAML workflows and batched commands
  • Controls Electron apps through CDP
  • Free and MIT licensed

Limits

  • The project is no longer maintained
  • Requires explicit OS accessibility permissions

Who it suits

agent-cu is aimed at AI agents, local models, and people who want to control desktop apps through a CLI. Users should be comfortable granting operating-system accessibility permissions and considering the project’s unmaintained status.

Verdict

agent-cu offers a broad set of local desktop-control commands and integrations. Its unmaintained status is a significant consideration for anyone choosing it for ongoing use.

Where Specifiction places it

Alternatives to agent-cuAll alternatives →