Aster Universal MCP

Aster Universal MCP is an open-source, donationware project by Tony Yustein that connects compatible AI assistants such as Claude or ChatGPT to a computer the user owns. The current release is v0.1.0b3, a manual technical beta with separate Windows and Linux source archives. Access uses OAuth 2.0 with PKCE pairing plus local administrator approval of each client.

What It Is

Aster is a self-hosted Model Context Protocol (MCP) gateway that runs on Windows or Linux. An AI client connects to its HTTPS MCP address, and authenticated tool requests pass through a tunnel to the local gateway. The AI model itself still runs with its provider. Once a client is approved, the assistant can work with real files, run commands and manage background jobs on the host.

How the Workflow Operates

The owner picks the machine, pairs the client over OAuth, and then approves that exact client ID locally from an administrator terminal. A remote request cannot approve itself. The connection_status tool reports the client's approval status, permissions and workspace. Setup counts as complete only after harmless native commands (hostname, whoami or id) return a completed status with exit code 0.

Documented tools include permission-controlled command execution (execute_command on Windows, run_command on Linux), directory listing, background jobs that can be inspected and cancelled, and chunked file transfer verified with SHA-256. The documentation describes a Cloudflare Tunnel as a way to publish a stable HTTPS hostname without router port forwarding.

Security Model and Limits

The project describes itself as a privileged administration beta, and recommends starting on a disposable machine. Commands inherit the gateway account's privileges on Windows and run as root on Linux; the workspace is not a sandbox. Startup is manual, with no installer, service or autostart. Qualified reference setups are Windows 11 with Python 3.11 x64 and Ubuntu 24.04 x86_64 with Python 3.12; macOS, WSL and ARM are not qualified. Multi-assistant collaboration, memory integrations and guided server setup are listed as roadmap goals, not current features.