Analysis updated 2026-06-24 · repo last pushed 2026-05-22
Run an AI coding agent on a local project folder without sending code to a SaaS.
Connect a teammate's machine to a remote OpenCode server using OpenWork client mode.
Save a common refactor or scaffolding workflow as a reusable Skill in .opencode/skills.
Approve or deny privileged agent actions interactively from a desktop permissions panel.
| different-ai/openwork | vscodevim/vim | jupyterlab/jupyterlab | |
|---|---|---|---|
| Stars | 15,157 | 15,147 | 15,142 |
| Language | TypeScript | TypeScript | TypeScript |
| Last pushed | 2026-05-22 | 2026-05-21 | — |
| Maintenance | Maintained | Maintained | — |
| Setup difficulty | moderate | easy | easy |
| Complexity | 3/5 | 2/5 | 3/5 |
| Audience | vibe coder | developer | data |
Figures from each repo's GitHub metadata at analysis time.
Pre-built binaries for macOS and Linux only, Windows requires the paid plan, and dev builds need Bun, Rust, pnpm, and WebKitGTK.
OpenWork is an open-source desktop application that lets you run AI coding agents on your own computer. The README positions it as the open-source alternative to Claude Cowork and Codex, two commercial desktop products in the same category. The actual AI agent that does the work underneath is a separate project called OpenCode, OpenWork is a polished interface and orchestration layer around it. Everything OpenCode can do is available through OpenWork, with the option to fall back to the OpenCode command-line tool whenever you want. The stated philosophy is local-first but cloud-ready. You install OpenWork on your laptop, click once, and start sending prompts to an agent that runs on your own machine, against a project folder you pick. If you later want to connect to a remote server, or share a session with teammates, that is an explicit opt-in step rather than the default. The README pitches OpenWork as a way to package agentic workflows so a team can run the same repeatable, productized process again and again. The feature list includes a host mode where OpenCode runs locally and a client mode that connects to a remote OpenCode server by URL. You can create chat sessions and send prompts. A live stream over Server-Sent Events shows real-time progress. An execution-plan view renders the agent's todo list as a timeline. A permissions panel surfaces requests from the agent for privileged actions, with allow-once, allow-always, or deny. Templates store common workflows locally, and a Skills manager handles the OpenCode plugins in a .opencode/skills folder. The project is built with Tauri, which means a Rust shell wrapping a webview, with the user interface written in TypeScript. Getting it running for development needs Node.js with pnpm, the Rust toolchain, Bun 1.3.9 or newer, Xcode command-line tools on macOS, and WebKitGTK 4.1 development packages on Linux. Pre-built downloads exist for macOS and Linux. Windows access is currently only available through the paid support plan. A separate npm package called openwork-orchestrator provides a command-line host that runs the same backend without the desktop UI. The repository is also the source of a commercial offering. The README links to an enterprise plan with SSO, SLA support, long-term support versions, and feature prioritisation, plus a hosted OpenWork Cloud workers product. By default, host mode binds only to 127.0.0.1, and model reasoning is hidden in the UI. The full README is longer than what was shown.
OpenWork is an open-source desktop app that runs AI coding agents locally on your machine. It wraps the OpenCode engine in a Tauri UI with sessions, plans, and permissions.
Mainly TypeScript. The stack also includes Tauri, Rust, TypeScript.
Maintained — commit in last 6 months (last push 2026-05-22).
License not stated in the README, check the LICENSE file in the repo before redistributing.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
Verify against the repo before relying on details.