Open Computer Use
Give your coding agent isolated Docker sandboxes with bash, browser, docs, and sub-agents so it can execute multi-step tasks like a dedicated computer.
Overview
Open Computer Use is an MCP server for the Build phase that gives LLMs Docker sandboxes with bash, browser, docs, and sub-agents for autonomous computer use.
What is this MCP server?
- Docker-based sandboxes that act as a full computer environment for LLMs
- Bash, browser, documentation access, and sub-agent orchestration in one MCP server
- Model-agnostic positioning: bring your own LLM to the sandbox
- Open Computer Use v0.8.12.6 from Yambr GitHub repository
- Fits advanced agent loops that go beyond read-only repo edits
- Server version 0.8.12.6
- Capabilities listed: Docker sandboxes, bash, browser, docs, sub-agents
Community signal: 92 GitHub stars.
What problem does it solve?
Agent workflows stall when the model cannot run shell commands, browse, or spawn helpers in an isolated environment next to your editor.
Who is it for?
Advanced indie agent builders prototyping computer-use features, eval harnesses, or internal dev automation with Docker on their machine.
Skip if: Beginners who only need snippet completion, or teams that require fully managed cloud sandboxes with enterprise compliance out of the box.
What do I get? / Deliverables
After you deploy the server and Docker stack, your agent can execute multi-step computer-use tasks inside sandboxes instead of only suggesting commands.
- Runnable isolated environments for agent-driven bash and browser tasks
- Sub-agent delegation patterns for longer autonomous workflows
Recommended MCP Servers
Journey fit
Build agent-tooling is the canonical shelf because you adopt this server while wiring autonomous execution into your product or internal agent stack. Agent-tooling matches sandbox provisioning and tool surfaces—not launch SEO or grow analytics.
How it compares
MCP sandbox runtime for LLM computer use, not a single-purpose integration like GitHub or a skills.sh methodology skill.
Common Questions / FAQ
Who is Open Computer Use for?
It is for developers building or testing agentic systems who want Docker-isolated bash, browser, and sub-agent tooling via MCP.
When should I use Open Computer Use?
Use it during build agent-tooling when your workflow needs executed commands and browsing, and during ship testing when you validate agent behavior in sandboxes.
How do I add Open Computer Use to my agent?
Follow the Yambr/open-computer-use GitHub setup to run the MCP server with Docker available, then register the server in Claude Code, Cursor, or your MCP host per their docs.