Sandbox
@iFurySt/open-codex-computer-use

MCP computer-use service for desktop automation

open-computer-use provides a local Computer Use runtime that agents reach through MCP. It runs on macOS, Linux, and Windows, with macOS accessibility and screen recording permissions required for desktop control. The repo also includes install commands for Codex, Claude Code, Gemini CLI, and opencode, plus a skill package and plugin support for Codex App.

1,996 stars179 forksSwiftUpdated 7d ago
Open (Codex) Computer Use
lambda lambda5.7k views • 5 months ago

Videos about this repo

Who it's for

Builders who want their agent to click, type, and move through desktop apps through MCP.

What it delivers

You can give an agent controlled access to desktop apps without building a custom computer-use layer yourself.

What it does

MCP computer use server

Exposes Computer Use actions through an MCP interface that any compatible agent or client can call.

Cross-platform desktop control

Runs on macOS, Linux, and Windows, with macOS-specific accessibility and screen recording setup.

Agent install commands

Provides built-in commands to install the service into Codex, Claude Code, Gemini CLI, and opencode configs.

Single-call and sequence execution

Supports one-off tool calls and multi-call sequences that reuse state across steps.

Skill package

Includes an installable `open-computer-use` skill for Codex and Claude Code.

Validation and smoke tests

Includes smoke, stress, and agent test scripts for checking the service from a source checkout.

How to get it

  1. 1Run
    npm i -g open-computer-use
  2. 2On macOS, run it once and grant Accessibility and Screen Recording. Windows and Linux do…
    open-computer-use
    # or
    ocu
  3. 3Before using it, install it into your agent
    # Install into Codex by writing to ~/.codex/config.toml
    open-computer-use install-codex-mcp
  4. 4Install the skill directly
    # Install for Codex
    npx skills add iFurySt/open-codex-computer-use -g -a codex --skill open-computer-use -y
    npx skills ls -g -a codex | rg 'open-computer-use'
  5. 5Install for Claude Code
    npx skills add iFurySt/open-codex-computer-use -g -a claude-code --skill open-computer-use -y
  6. 6Update an existing global install, including the Codex install created above
    npx skills update open-computer-use -g -y

README

open-computer-use

English 简体中文 Release Ask DeepWiki LLMAPIS

[!TIP] Interested in Browser Use? Check out open-browser-use.


open-computer-use is an open-source Computer Use service wrapped as MCP. Any AI agent or MCP client can use it to run Computer Use on macOS, Linux, and Windows.

This project was inspired by OpenAI's Codex Computer Use. It showed that non-intrusive CUA can be built on top of Accessibility, so I decided to build an open-source version.

I started this repo with my harness template, a template for quickly spinning up AI-first projects. It has been one of our most useful workflows lately, especially for nearly 100% AI-generated projects. I also wrote a post about the methodology behind it.

Demos

Codex App and Codex CLI

Open Computer Use custom demo cover

open-computer-use used as Computer Use in Codex App and Codex CLI, matching the official experience.

Gemini CLI

https://github.com/user-attachments/assets/eacb3b15-f939-46c7-b3b3-6f876977a58d

Gemini CLI connects to open-computer-use through MCP and runs full Computer Use actions.

Linux

https://github.com/user-attachments/assets/e036b1c8-2200-4896-abd4-19225915cf66

open-computer-use running on Linux.

Quick Start

npm i -g open-computer-use

The npm package also exposes ocu as the short CLI alias.

[!IMPORTANT] The macOS runtime requires macOS 14.0 or later.

On macOS, run it once and grant Accessibility and Screen Recording. Windows and Linux do not need this step.

open-computer-use
# or
ocu

Before using it, install it into your agent:

# Install into Codex by writing to ~/.codex/config.toml
open-computer-use install-codex-mcp

Or add it to your own client manually:

{
  "mcpServers": {
    "open-computer-use": {
      "command": "open-computer-use",
      "args": ["mcp"]
    }
  }
}

Skill

Install the skill directly:

# Install for Codex
npx skills add iFurySt/open-codex-computer-use -g -a codex --skill open-computer-use -y
npx skills ls -g -a codex | rg 'open-computer-use'

Install for Claude Code:

npx skills add iFurySt/open-codex-computer-use -g -a claude-code --skill open-computer-use -y

Update an existing global install, including the Codex install created above:

npx skills update open-computer-use -g -y

You can also manually download and install the open-computer-use skill.

More

Besides the MCP JSON config above, you can also use the built-in commands:

# Install into Codex by writing to ~/.codex/config.toml
open-computer-use install-codex-mcp
ocu install-codex-mcp

# Install as a Codex plugin, mainly for Codex App
open-computer-use install-codex-plugin

# Install into Claude Code by writing to ~/.claude.json
open-computer-use install-claude-mcp

# Install into Gemini CLI for the current project by writing to ./.gemini/settings.json
open-computer-use install-gemini-mcp

# Install into Gemini CLI user config instead
open-computer-use install-gemini-mcp --scope user

# Install into opencode by writing to ~/.config/opencode/opencode.json (or the active config file)
open-computer-use install-opencode-mcp

# Call a single Computer Use tool and print the MCP-style JSON result
open-computer-use call list_apps
ocu call list_apps
open-computer-use call get_app_state --args '{"app":"TextEdit"}'

# Run a sequence in one process so element_index state can be reused
# Sequence runs sleep 1s between successful operations by default
open-computer-use call --calls '[{"tool":"get_app_state","args":{"app":"TextEdit"}},{"tool":"press_key","args":{"app":"TextEdit","key":"Return"}}]'
open-computer-use call --calls-file examples/textedit-overlay-seq.json --sleep 0.5

# Check permissions; onboarding only opens when something is missing
open-computer-use doctor

# Run local validation from a source checkout
make smoke
OPEN_COMPUTER_USE_STRESS_LOOPS=20 make stress
make agent-smoke
make agent-smoke SCENARIO=fixture-full
node ./scripts/run-agent-smoke-tests.mjs --agents=claude,codex --command=open-computer-use
node ./scripts/run-agent-smoke-tests.mjs --scenario=fixture --agents=claude,codex --command=open-computer-use
node ./scripts/run-agent-smoke-tests.mjs --scenario=fixture-full --agents=claude,codex --command=open-computer-use
OPEN_COMPUTER_USE_HERMES_PROVIDER=anthropic OPEN_COMPUTER_USE_HERMES_MODEL=claude-opus-4-20250514 make agent-smoke AGENTS=hermes SCENARIO=fixture-full
node ./scripts/run-agent-smoke-tests.mjs --agents=hermes --hermes-provider=anthropic --hermes-model=claude-opus-4-20250514
node ./scripts/run-agent-smoke-tests.mjs --scenario=fixture --agents=hermes --hermes-provider=anthropic --hermes-model=claude-opus-4-20250514
node ./scripts/run-agent-smoke-tests.mjs --scenario=fixture-full --agents=hermes --hermes-provider=anthropic --hermes-model=claude-opus-4-20250514 --hermes-max-turns=12

# Show help
open-computer-use -h
ocu -h

Cursor Motion

Cursor Motion is an open-source cursor motion system for macOS, based on public information shared by members of the Software.Inc team. You can download the app from the Releases page.

Cursor Motion custom demo cover

Star History

Star History Chart for open-computer-useStar History Chart for open-computer-use

License

MIT. Third-party attribution is listed in THIRD_PARTY_NOTICES.md.

Files in the repo

Repository payload26 top-level entries
  • .agents
  • .github
  • apps
  • artifacts
  • assets
  • docs
  • examples
  • experiments
  • packages
  • plugins
  • scripts
  • skills
  • .gitattributes
  • .gitignore
  • AGENTS.md
  • CLAUDE.md
  • CODEOWNERS
  • CONTRIBUTING.md
  • LICENSE
  • Makefile
  • package.json
  • Package.swift
  • README.md
  • README.zh-CN.md
  • SECURITY.md
  • THIRD_PARTY_NOTICES.md

Discussion (0)

Ask about usage, or say what you built with it

Sign in to join the discussion.

No comments yet. Be the first to say what this is good for.

More connectors

High-performance code intelligence MCP server. Indexes codebases into a persistent knowledge graph — average repo in milliseconds. 158 languages, sub-ms queries, 99% fewer tokens. Single static binary, zero dependencies.

43k

Universal provider proxy for OpenAI Codex & Claude Code — use any LLM (Claude, Gemini, Grok, DeepSeek, Ollama…) with Codex CLI, App, SDK, and Claude Code

14k
okf-memory/
okf-agent-memory

Git-native persistent memory for AI coding agents. Implements Google OKF v0.2 with sub-300µs in-memory BM25 search, embedded MCP server, and progressive disclosure. Slashes token bloat by 80% with zero external databases or dependencies. Built in pure Go.

547
tirth8205/
code-review-graph

Local-first code intelligence graph for MCP and CLI. Builds a persistent map of your codebase so AI coding tools read only what matters, with benchmarked context reductions on reviews and large-repo workflows.

31k
2akouwu/
reverify

Stop your AI from making things up — it proposes, deterministic tools decide, every claim checked against ground truth with evidence. Grounded facts and context survive resets. Reverse engineering is the proving ground. MCP server + CLI.

1.1k
t8y2/dbxConnectors

20 MB lightweight cross-platform database client for 90+ databases, including MySQL, PostgreSQL, SQLite, Redis, MongoDB, DuckDB, SQL Server, and Dameng. Built-in AI, MCP Server, CLI, desktop and Docker. | 轻量级跨平台数据库管理工具,支持 MySQL、PostgreSQL、SQLite、Redis、MongoDB、达梦等 90+ 数据库,提供桌面端、Docker、CLI、内置 AI 助手和 MCP Server。

19k