Panel

Assistant

Connect an AI client to your server so it can run panel actions for you.

The Assistant page lets you connect your own AI to your server. Once connected, you can ask it in plain language to start the server, run a console command, edit a config file, or set up a scheduled task, and it performs the action in the panel for you.

The connection uses the Model Context Protocol (MCP), an open standard that AI applications use to talk to external tools. The panel acts as an MCP server for one game server, and your AI client acts as the MCP client.

In-panel chat is not available yet

The first card on the page shows a chat preview with a Locked badge. Chatting directly inside the panel is coming soon. For now, the Assistant page is where you connect an outside AI client, and the conversation happens in that client.

What You Need

An AI client that supports MCP. The panel lists the ones people use most:

  • Claude Desktop
  • Claude Code
  • Cursor
  • ChatGPT
  • Any MCP client

You also need the AI/MCP permission on the server. If you own the server you already have it. If you were invited as a collaborator, see Permissions below.

Connecting Your AI

The Connect your AI card handles the whole setup. You never have to write a configuration file by hand.

  1. Open the Assistant tab for your server.
  2. Click Connect your AI.
  3. Read the warning, then click I understand, connect. The panel creates an access token for you.
  4. A block of text appears. Click it to copy it.
  5. Paste that block into your AI client and send it. It contains the endpoint, the token, and ready made configuration for both app based and command line clients, so the AI sets itself up from there.
  6. Click Done, I've saved it to close the dialog.
  7. Ask your AI to list its available tools. If it can name tools such as power or send_command, the connection works.

The card also shows the MCP Endpoint, which ends in /api/mcp. Click it to copy it. You only need it if you prefer to configure your client manually, since the pasted block already includes it.

Once a token exists, the card shows a green Connected badge, the identifier of your token, and buttons to Reconnect or Revoke.

Your Token

The token is what proves to the panel that the AI is acting on your behalf. A few rules to keep in mind:

  • It is shown only once. The panel never displays it again after you close the dialog. If you lose it, click Reconnect to get a new one.
  • You get one token per server. Connecting again replaces the previous token, so any AI still using the old one immediately loses access.
  • It is tied to a single server. A token created here cannot see or touch any other server on your account.
  • It does not expire. It stays valid until you revoke it or the server is deleted.
  • It is personal. Each person with access to the server creates their own token, and you only ever see and revoke your own.

Click Revoke at any time to cut the connection. Any AI client using that token loses access to the server right away.

Treat the token like a password

Anyone who has your token gets full control of this server through their own AI. Never share it, post it in a Discord message or a support ticket, or commit it into code. If you think it leaked, revoke it immediately and connect again to get a fresh one.

What an AI Can Do

The Exposed Tools card lists the categories a connected AI can act on. The list adapts to your own permissions, so you only ever see what you are allowed to do yourself.

  • Power: start, stop, restart, and force kill the server, and read its current state, resource usage, and limits.
  • Console: send commands to the server and read recent console output.
  • Files: browse directories, read and write files, create folders, rename, copy, delete, compress and extract archives, change file permissions, and download a file from a URL into the server.
  • Schedules: list, create, edit, run, and delete schedules, and add or remove their tasks, including command, power, and backup tasks.
  • Network: list allocations, add one, change the primary allocation, and delete one.
  • Startup: read the startup variables and change their values.
  • Databases: list, create, and delete databases, on applications that support them.

Because an AI works through the same systems you do, the usual limits still apply. It cannot add an allocation past your plan limit, delete your last allocation, or change a port that a startup variable manages for you.

What an AI Can Never Do

Some actions are blocked at the panel level, no matter what permissions you hold:

  • Deleting backups. This is refused for every token.
  • Reading or managing collaborators. An AI cannot list, invite, edit, or remove the people who have access to your server.

An AI also cannot create, restore, or list backups directly. The only way it can help with backups is by creating a scheduled backup task for you, which then runs on its own schedule. To take or restore a backup right now, use the Backups page yourself.

Permissions

A token can never do more than the person who created it. Every request is checked against your current permissions on the server, so the tools your AI sees are filtered to match your access.

Those checks happen live, on every single action. If a server owner reduces your access or removes you from the server, your token narrows or stops working immediately, with no need to revoke anything.

Collaborators receive the AI/MCP permission by default when they are invited, which lets them create and revoke their own token. See Collaborators for how server access is shared.

The Assistant tab appears greyed out in the sidebar when you do not hold the AI/MCP permission on that server. It is also unavailable while a server is suspended or under maintenance.

Activity Logging

Nothing an AI does is hidden. Every action it takes is written to your server Activity log exactly like an action you performed by hand, so you can always see what was run, and when. Creating and revoking a token is logged too.

Troubleshooting

Table of Contents