Wolfpack: Mobile & Desktop AI Agent Command Center

What changed

Wolfpack has been released as a self-hosted control room designed for AI coding agents, enabling remote monitoring and control from both desktop and mobile devices. The system allows users to maintain persistent agent terminals on their own machines and access them even when away from their desks. It supports a range of built-in AI coding agent providers, including Claude Code, Codex, and Gemini CLI, with Shell as a fallback option. Custom commands and wrappers on the system's PATH are also supported. Sessions are managed by a Rust PTY broker, ensuring that terminals remain active even if the web server restarts or the browser is closed. The project emphasizes security through private Tailscale networking, advising users to configure device/user ACLs and consider JWT for shared Tailnets. Installation can be done via a curl installer for a persistent CLI or using package runners like Bunx or npm. The setup process involves choosing a projects directory, port, Tailscale remote access configuration, and optional Pi integration. Success is indicated by a working local URL, a verified Tailnet HTTPS URL with a QR code for remote access, and a clean wolfpack doctor diagnosis. Uninstalling is straightforward with specific commands for each installation method.

Why it matters for builders

Wolfpack offers AI builders a significant advantage by decoupling agent sessions from a specific machine or active browser window. This means developers can initiate a task on their desktop, leave their workstation, and then monitor or interact with the agent's progress from their phone via the Progressive Web App (PWA). The ability to see which agents require input and to open live terminals remotely reduces context switching and ensures that agent tasks are not stalled due to a developer's physical absence. Furthermore, the integration with Tailscale provides a secure, private connection without relying on third-party hosted relays, giving builders more control over their development environment.

Practical impact

Developers can leverage Wolfpack to manage their AI coding agent workflows more efficiently. The mobile PWA allows for on-the-go monitoring and intervention, turning a smartphone into a functional control surface for AI development. This is particularly useful for tasks that require occasional input or oversight. The desktop interface provides a visual control room with live session previews, consolidating management of multiple agents. Builders can use the wolfpack doctor command to verify their installation and configuration. For advanced automation, the command-line interface allows for programmatic session creation and agent spawning, enabling integration into larger CI/CD pipelines or custom agent orchestration systems. The project also supports agent skills for Pi, allowing for more sophisticated task delegation and durable routing.

Caveats and source limits

The provided source material details the functionality and installation of Wolfpack but does not include specific benchmark results or performance metrics. Information regarding pricing or commercial support is also absent, as it is a self-hosted, open-source project. While the project mentions security features like Tailscale integration and JWT considerations, a comprehensive security audit report is not provided. The latest release version mentioned is v1.6.3, but details on the release date of this specific version are not available beyond the general project publication date. The excerpt also notes "3 AI signals, 7 developer signals" which are not further elaborated upon in the provided text.