wheeljack documentation
Run terminals, coding agents, plans, and review evidence in one local-first desktop workspace.
wheeljack is a desktop terminal multiplexer for Windows and macOS. It combines recursive terminal splits, structured coding-agent sessions, project planning, isolated task lanes, reusable Bots, and review evidence without moving project or session state into a hosted account.
- Start a workspace. Install wheeljack, open a project folder, and start a shell or coding agent in a few focused steps.
- Connect your agents. Use Claude Code, Codex CLI, OpenCode, Pi, or a plain shell. Existing CLI authentication stays outside wheeljack.
- Plan and review. Turn project intent into task contracts, isolated Git lanes, verification, and inspectable review evidence.
- Stay local-first. Projects, layouts, preferences, transcripts, and Plan state remain in local SQLite storage on your machine.
Choose a path
Section titled “Choose a path”- New to wheeljack: install the app, then open your first project.
- Bringing a coding-agent CLI: connect and verify an agent.
- Running coordinated work: learn the Plan and review workflow.
- Evaluating the trust boundary: read local data and permissions.
- Extending wheeljack: start with architecture and the agent adapter contract.
Supported platforms
Section titled “Supported platforms”Release builds target Windows x64 and universal macOS. Linux is not a supported release target yet. Download the latest packages from GitHub Releases.
wheeljack does not bundle model access, agent subscriptions, or provider credentials. Each coding-agent CLI uses its own account, provider, limits, and network connection.