Included with every server

Point your AI at your game server

Connect your own AI assistant and ask it real questions about your server. Why did it crash. What is eating the disk. Which addon is throwing errors. It reads the actual files and logs, not a guess.

One command in your terminal

claude mcp add --transport http --scope user zeros-host https://panel.zerosrp.co.uk/mcp --header "Authorization: Bearer YOUR_API_KEY"

Create your key in the panel under Account, then Connect your AI, where you can also limit it to certain servers and actions. Paste it in place of YOUR_API_KEY, then restart your assistant.

Then check it saved the key: claude mcp get zeros-host should show an Authorization header. Do not rely on it reporting “Connected”. Older Claude Code versions said that while silently dropping the key. If the header is missing, update Claude Code.

Works with your assistant

Anything that speaks the Model Context Protocol can connect. The panel builds the exact snippet for whichever one you use, with your key already in it.

Set it up in your panel

Connects directly

Paste one command or config block and you are connected.

Claude CodeAnthropic

CursorAnysphere

VS CodeMicrosoft

GitHub CopilotGitHub

WindsurfCodeium

ClineVS Code extension

ContinueContinue Dev

Gemini CLIGoogle

GOGooseBlock

Needs a one line bridge

Needs a small local relay, because these expect a local process rather than a URL.

COCodex CLIOpenAI

ZedZed Industries

JetBrains AIJetBrains

Coming soon

Hosted assistants that require OAuth Dynamic Client Registration. We are building it.

Claude web and desktopAnthropic

CHChatGPTOpenAI

Listed by what actually works today rather than what we would like to claim. When the hosted assistants gain support, they move up the list and this page changes with them.

Things people actually ask it

You talk to your assistant normally. It decides which of the tools below to use.

“Read the last 50 console lines and tell me why the server crashed”

“What is using the most disk space on my server”

“Search my addons folder for anything that looks like a backdoor”

“Which of my servers is closest to its memory limit”

“Show me what changed on this server in the last day”

“Has my AI touched anything on this server”

Everything it can see

Eleven read tools. Writing files and power actions like restart are rolling out now in an opt-in pilot, behind per-action confirmation.

Your servers

list_servers

Every server on your account. Start here: it returns the identifier the other tools need.

server_details

Game, limits, allocations and current state for one server.

resource_usage

Live CPU, memory, disk and network use.

Files

list_directory

What is in a folder.

list_directory_recursive

A whole subtree in one call, for questions like "what addons are installed".

read_file

The contents of a config, log or script.

search_files

Find files by name or content, with optional subtree, extension, regex and result limits.

storage_breakdown

How much of your disk and backup allowance is used.

Logs and history

console_logs

Recent console output, with colour codes stripped so it reads cleanly.

activity_log

What happened on this server, for "what changed just before it broke".

api_activity

What your AI has been doing, separated from your own sign-ins.

What it cannot do, and why

Handing an AI access to a live game server deserves more than a promise, so here is exactly where the limits are.

You choose exactly what it can do

A new key is read-only by default: it can look, never touch. Writing files and power actions like restart are separate permissions you grant per key, per server, so nothing on your server can change unless you explicitly allowed it when you made the key.

Your settings secrets are stripped out

Your server settings include things like a Discord bot token, a Steam login token or an RCON password. Those values are removed before anything is sent, because tool output travels to whichever AI provider you use. You see the setting names, never their values. Files that exist to hold credentials, like .env, secrets files and private keys, are refused outright, and we scan file contents and logs for things that look like keys and webhooks. That last part is pattern matching rather than a promise: if you have pasted a password into an addon file, treat it as readable and move it into your server settings.

Your key, your permissions

We hold no credential of our own. Your panel API key is passed straight through, so the AI can reach exactly the servers you can and nothing else. Revoke the key and access stops immediately.

Server content is treated as data

Your logs contain text typed by players. Everything returned is labelled to the AI as information to analyse rather than instructions to follow, so a player writing something that looks like a command cannot steer your assistant.

It can

Read any file on your server
Read console output and logs
See resource use and limits
See your account audit trail

It cannot

Edit, create or delete files
Start, stop or restart anything
Run console or RCON commands
See passwords, tokens or keys
Touch anyone else’s server

See everything your AI has done

Every action taken with an API key is recorded and marked as coming from a key rather than from you signing in. That includes file reads, so you can see what an assistant looked at, not just what it changed. Ask it to check api_activity, or open Activity in your panel account at any time.

“Show me every file my AI has read on this server today”

Reads are logged the same way writes are, so as write access rolls out there is already a full record of everything an assistant touches.

Questions

Which AI assistants work with this?

Claude Code works today with one command. The rest land as they finish MCP support. The list on this page updates when they do.

Can the AI break my server?

Only if you let it, and even then it asks first. A key is read-only unless you grant writing or power actions. When you do, every change is a two-step confirmed action: the AI proposes the exact change and nothing happens until it is confirmed, and files like .env, secrets and keys can never be written no matter what a key is granted.

Can it see my passwords and tokens?

No. Values that look like secrets (bot tokens, Steam login tokens, RCON passwords) are stripped before anything leaves the panel. The AI sees that a value is set, not what it is.

How do I see what the AI has done?

Ask it to check api_activity, or open Activity in your panel account. Every API-key action is marked, including file reads.

How do I turn it off?

Delete the key under Account → Connect your AI. Access stops instantly. You can also narrow a key: limit it to specific servers, or block file and log reads, instead of deleting it.

Does this cost anything?

No. Included with every server; you only pay whatever your AI assistant costs.

Already have a server with us

Create an API key in your panel and run the command above. If you do not have a server yet, every plan includes this, and so does the free trial.

Create an API key See plans