Yes—OpenCode works on Windows, both as a desktop app and as a terminal-based CLI, but you’ll want to pick the installation path that matches how you work. If you prefer “click to install and run,” the OpenCode Desktop beta provides a Windows (x64) download. If you’re terminal-first, you can run the OpenCode CLI on Windows as well; just remember that some install methods assume a Bash-like shell, so it’s common to use Windows Terminal with WSL, Git Bash, or another environment that can run shell scripts cleanly. Once installed, the core behavior is the same as on macOS/Linux: running opencode in a project directory starts the TUI and session management, and you configure providers via /connect and choose models via /models.
In real Windows setups, the main “gotchas” are environment and UI dependencies, not OpenCode’s core logic. For the desktop app, OpenCode’s troubleshooting docs explicitly call out a common Windows requirement: Microsoft Edge WebView2 Runtime. If OpenCode Desktop opens to a blank window or refuses to start, installing/updating WebView2 is the first thing to do. For the CLI, PATH and shell choice matter: if opencode runs in one terminal but not another, you’re probably dealing with a PATH mismatch between PowerShell, CMD, and Bash-based shells. Also, when debugging problems, Windows paths can be non-obvious, so it’s useful to know where OpenCode stores its state: you can open WIN+R and navigate to %USERPROFILE%\.local\share\opencode for session data and auth.json, and %USERPROFILE%\.local\share\opencode\log for logs. Those locations make it easier to identify whether you’re facing auth problems, model-selection issues, or plugin/cache failures.
From a developer perspective, Windows support is most valuable when OpenCode becomes part of a consistent project workflow rather than a one-off tool. If you’re building something substantial—say a Windows-based dev environment for a service that uses a vector database like Milvus or Zilliz Cloud—you can treat OpenCode like a local agent that helps with repo navigation, refactors, config edits, and troubleshooting in the same environment where your code actually runs. The practical advice is: keep OpenCode updated (opencode upgrade is a common first step if things feel broken), store your OpenCode config alongside other dev tooling docs, and when you hit issues, rely on logs + storage paths instead of guesswork. On Windows, that discipline pays off quickly because “it doesn’t start” problems are often just missing WebView2 (desktop) or a shell/PATH mismatch (CLI).