MCP Servers · iFurySt

open-codex-computer-use

open-codex-computer-use is an open-source desktop automation service that acts as a computer use agent, allowing AI models to control macOS, Linux, and Windows via an accessibility-based interface. It wraps this functionality as an MCP (Model Context Protocol) server, enabling integration with Claude, Codex, Gemini, and other AI agents.

Talk to DEV.co Discuss Implementation

Source: GitHub — github.com/iFurySt/open-codex-computer-use

1.4k

GitHub stars

133

Forks

Swift

Primary language

MIT

License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

Field	Value
Repository	iFurySt/open-codex-computer-use
Owner	iFurySt
Primary language	Swift
License	MIT — OSI-approved
Stars	1.4k
Forks	133
Open issues	8
Latest release	v0.1.54 (2026-06-26)
Last updated	2026-07-08
Source	https://github.com/iFurySt/open-codex-computer-use

What open-codex-computer-use is

Built in Swift/Node.js, the project exposes computer control primitives (app state, clicking, keyboard input) through an MCP interface without intrusive system hooks. It leverages accessibility APIs on macOS and Linux, with platform-specific implementations for Windows, and can be invoked as a standalone CLI, npm package, or MCP server.

Quickstart

Get the open-codex-computer-use source

Clone the repository and explore it locally.

terminalbash

1git clone https://github.com/iFurySt/open-codex-computer-use.git2cd open-codex-computer-use3# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

AI-driven GUI automation workflows

Enable Claude, Codex, or Gemini to autonomously interact with desktop applications, file systems, and GUI elements for end-to-end task completion (e.g., form filling, data entry, cross-app workflows).

Development tool augmentation

Integrate with AI code assistants to automate repetitive development tasks such as testing frameworks, IDE navigation, and simulator/emulator interaction on local machines.

Cross-platform accessibility automation

Provide accessibility-focused desktop control on macOS, Linux, and Windows through a unified MCP API, supporting accessibility-first automation design patterns.

Implementation considerations

Requires explicit grant of Accessibility and Screen Recording permissions on macOS; user must manually approve in System Preferences before first use.
State management is retained across sequential calls (element_index reuse); sequences run with 1s sleep by default between operations—tune for your AI agent's decision cycle.
Primary language is Swift; Node.js wrapping is used for CLI/npm. Review build and runtime dependencies for your target platform (macOS vs. Linux vs. Windows).
MCP server pattern allows multi-client use, but concurrent AI agent access to the same desktop session may cause race conditions or unpredictable UI state transitions.
Test permissioning and platform-specific API coverage (Linux via AT-SPI, Windows via UIA) early; not all UI frameworks expose equivalent accessibility hooks.

When to avoid it — and what to weigh

Server-side or headless automation needs — This tool requires a graphical desktop environment and accessibility APIs. It is not suitable for headless servers, containers, or cloud environments without a GUI.
Security-critical control of sensitive systems — Runs user-level processes with screen capture and input simulation. Not recommended for controlling financial systems, healthcare platforms, or other high-risk systems without additional isolation and audit controls.
Performance-critical, sub-millisecond interactions — GUI automation inherently involves polling, network latency (if MCP is remote), and OS-level delays. Not suitable for real-time or latency-sensitive operations.
Closed or proprietary application control — Accessibility APIs may not work consistently on all proprietary or heavily sandboxed applications. Test integration with target apps before relying on it.

License & commercial use

MIT License. Permissive OSI-approved license permitting commercial use, modification, and redistribution with attribution and inclusion of the license notice.

MIT permits commercial use, but review liability disclaimers in the license carefully. The project is provided 'as-is' without warranty. Any use for commercial automation should include error handling, monitoring, and audit trails; validate integration with proprietary or regulated systems independently.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

Signal	Assessment
Maintenance	Active
Documentation	Adequate
License clarity	Clear
Deployment complexity	Low
DEV.co fit	Good
Assessment confidence	High

Security considerations

Requires screen recording and input simulation permissions, granting broad visibility and control over the user's desktop session. No encryption, authentication, or audit logging mentioned. If used in shared or multi-tenant environments, or to control sensitive applications, implement external access controls, logging, and session isolation. Verify compliance with security policies before automating access to regulated systems.

Alternatives to consider

Anthropic's official Claude Computer Use (not open-source)

Direct integration with Claude API, likely more stable and optimized, but closed-source and proprietary. Use if API integration is preferred over local MCP deployment.

OpenAI's Codex Computer Use (proprietary)

Original inspiration; closed-source service. If you require official support and integration with OpenAI ecosystem, preferred over open-source alternative.

Robocorp RPA / UiPath

Enterprise RPA platforms with broader desktop and web automation, audit trails, and commercial support. Better for regulated industries and large-scale deployments, but higher cost and learning curve.

Software development agency

Build on open-codex-computer-use with DEV.co software developers

Evaluate open-codex-computer-use for your AI agent integration. Test permissions, platform compatibility, and MCP integration in a non-production environment first.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

mcp-use DesktopCommanderMCP Windows-MCP

Related on DEV.co

Explore the category and the services that help you build with it.

MCP Servers AI Development Custom Software Talk to DEV.co

open-codex-computer-use FAQ

Can I use open-computer-use without a desktop GUI?

No. This tool requires a graphical desktop environment with accessibility APIs (macOS Accessibility, Linux AT-SPI, Windows UIA). It cannot run on headless servers or in container environments without a display.

Does it work with web browsers or only native apps?

The README shows examples with Codex, Gemini, and TextEdit (native app). Web browser support depends on accessibility hooks in the browser and website; not explicitly tested or documented.

Can multiple AI agents control the same desktop simultaneously?

Not safely. The project is designed for single-user, single-agent workflows. Concurrent access from multiple agents will race condition on UI state; implement external queuing or locking if needed.

What is the latency between AI agent decision and screen capture?

Unknown; not benchmarked in documentation. Expect overhead from OS-level UI polling, accessibility API calls, and network latency if MCP runs on a different process. Test with your agent and target application.

Work with a software development agency

Adopting open-codex-computer-use is usually one piece of a larger software development effort. As a software development agency, DEV.co provides software development services and web development expertise — pairing senior software developers and web developers with your team to design, build, and operate mcp servers software in production.

Ready to automate desktop workflows with AI?

Evaluate open-codex-computer-use for your AI agent integration. Test permissions, platform compatibility, and MCP integration in a non-production environment first.

Talk to DEV.co Schedule a Consultation