Frequently asked questions

Short answers to the questions people ask before connecting an AI assistant to their machines.

What is WECOMMANDER?

WECOMMANDER is a remote MCP (Model Context Protocol) server that lets AI assistants such as Claude, ChatGPT and Cursor use the terminals and files of your own Windows, macOS and Linux machines through one MCP URL, with human approval for risky actions, per-device limits and a full audit log.

Which AI assistants work with it?

Any client that supports remote MCP servers: Claude (web, desktop, Claude Code), ChatGPT connectors, Cursor and others. Local mode also works with stdio-only clients.

Which operating systems are supported?

Windows 10 and 11, macOS 13 or later, and Linux (Ubuntu and Debian packages; other distributions with the binary).

Do I need to open a port or set up a VPN?

No. The agent makes an outbound TLS connection to the gateway; nothing listens on your machine.

Can the AI run any command on my server?

Only inside the folders you allow, as a normal user, and every command waits for your approval unless you allowed that command pattern or turned on autonomous mode for the device.

How do I approve actions quickly?

In the dashboard, from the approval e-mail, or with the Management Center desktop app on Windows, which shows each request as a notification with Approve and Deny buttons.

Does the AI see my passwords or API keys?

Common secrets in command and terminal output are masked on the device before the result is sent. File reads are exact, so keep secret files outside the allowed folders.

Can I undo a change the AI made?

File edits keep undo snapshots; the AI or you can restore the previous content if the file has not changed since.

Can I use it without an account?

Yes, in local mode: the agent serves MCP over stdio to a local client, with writes and commands off unless you enable them.

Can I host it myself?

Yes. The gateway runs with Docker Compose and PostgreSQL; the agent is the same.