MCP-wrapped Computer Use service
Runs as an open-source Computer Use service exposed through MCP, so agents and MCP clients can invoke it instead of using a closed vendor-specific integration.
Open Computer Use
Open Computer Use is an open-source Computer Use service wrapped as MCP for macOS, Linux, and Windows. It helps AI agents and MCP clients run desktop automation workflows through client setup commands or manual MCP configuration.
Open-source Computer Use for MCP clients
Open Computer Use is an open-source Computer Use service wrapped as MCP. It is designed so AI agents and MCP-capable clients can run Computer Use workflows on macOS, Linux, and Windows.
The repository positions the project as an open-source alternative to Codex Computer Use and says it was inspired by OpenAI’s implementation. The README also shows command-line setup, built-in install commands for several clients, and example usage for single actions, multi-step sequences, and local validation.
Core capabilities
Cross-platform desktop support
Supports macOS, Linux, and Windows according to the README, with a separate macOS permission step for Accessibility and Screen Recording.
Multiple client setup paths
Provides built-in install commands for Codex, Claude Code, Gemini CLI, and opencode, plus a manual `mcpServers` configuration example.
Single-action and sequenced runs
Includes a `call` command for one-off tool execution and a sequence mode that reuses element index state across multiple operations.
Permission and setup checks
Offers `doctor` to check permissions and onboarding only opens when something is missing, which helps users verify local setup before use.
Project validation commands
Lists local validation commands such as `make smoke`, `make stress`, and agent smoke-test scripts for exercising the project from a source checkout.
Practical use cases
MCP-based desktop automation
Use it when you want an AI agent to control a desktop app through MCP rather than through a proprietary Computer Use endpoint.
Cross-client agent workflows
Install it into Codex, Claude Code, Gemini CLI, or opencode when you want the same Computer Use backend available in different agent environments.
Local multi-step automation
Run it on your local machine to automate repeated GUI tasks, using the sequence mode when the same app state needs to be reused across steps.
Setup validation and debugging
Use the `doctor` command and local smoke tests to verify permissions and installation before putting it into a workflow.
Pros and Cons
Pros
- Open source and MIT licensed.
- Works across macOS, Linux, and Windows.
- Supports several MCP-capable clients, including Codex, Claude Code, Gemini CLI, and opencode.
- Documents both one-off tool calls and reusable multi-step sequences.
- Includes a permission check command and local smoke-test commands.
Cons
- The README provides limited detail on supported workflows beyond the documented commands and demos.
- Mac setup requires Accessibility and Screen Recording permission grants before use.
- The source does not list a formal hosted service or managed plan; it is presented as a self-hosted open-source tool.
FAQ
What is Open Computer Use?
It is an open-source Computer Use service wrapped as MCP. The README says any AI agent or MCP client can use it to run Computer Use on macOS, Linux, and Windows.
How do you set it up?
The README shows command-line installation with `npm i -g open-computer-use`, then installing it into an agent such as Codex, Claude Code, Gemini CLI, or opencode through the built-in MCP install commands or a manual MCP server entry.
Does it require special permissions?
The README says macOS users need to run it once and grant Accessibility and Screen Recording permissions. Windows and Linux do not need that step.
Which clients does it support?
The repository documents built-in commands for installing into Codex, Claude Code, Gemini CLI, and opencode, and it also includes a manual MCP configuration example. That indicates it is meant to work with MCP-capable clients rather than only one app.
Can it run multi-step Computer Use sequences?
The README includes a `call` command for running a single Computer Use tool and a `--calls` / `--calls-file` workflow for running sequences in one process so element index state can be reused.
Quick Facts
- Category
- Developer Tool
- Source domain
- github.com
- License
- MIT
- Platform support
- macOS, Linux, Windows
- Pricing
- Open-source self-hosted; no paid plan is listed in the repository
- Primary interface
- MCP client or command line
Open Computer Use Alternativen
Codex Plugins
Codex Plugins bundle reusable skills, app integrations, and MCP servers into workflows you can install in the Codex app or use from Codex CLI. They help extend Codex with connected-service tasks, reusable instructions, and shared team workflows.
AakarDev AI
AakarDev AI helps teams manage AI provider access, project-level setups, logs, and analytics from one dashboard. It supports BYOK workflows and lists providers including OpenAI, Google Gemini, Anthropic, Groq, Mistral AI, and Perplexity AI.
Arduino VENTUNO Q
Arduino VENTUNO Q is an edge AI computer for AI and robotics applications. It combines AI inference and deterministic control on a single board and is designed to work with Arduino App Lab.
Devin
Devin is an AI coding agent and software engineer that helps developers and engineering teams plan and execute complex software tasks. It is available through desktop, cloud, JetBrains, and CLI surfaces, with plans for individuals, teams, and enterprises.
ByteAsk
ByteAsk is a terminal-first AI coding agent for C and C++ that edits repositories and verifies changes with the real compiler, debugger, sanitizers, and tests before showing a diff. It offers a free tier plus paid plans, with editor connectors and zero-retention handling described in the source.
hob
hob is an independent workspace for coding agents that keeps agent sessions, terminals, history, and follow-up work organized around the tools and providers you already use. It is aimed at developers who want local control over routing, history, and workspace structure rather than a bundled model stack.