GuidesOnly show guides for your app
GuidesGuidesFelix can read your server and make changes for you — here's honestly what it can see, what it can do, and how the approval and trust controls keep you in charge.
Falix has a built-in AI assistant named Felix that can actually work on your server, not just chat about it. Point it at a server and it can read the files, logs, and console, diagnose a problem, and — with your approval — fix it: edit a config, install a plugin, restart the server. This guide is an honest look at what it can see, what it can do, and the controls that keep you in the driver's seat. Open it at /ai-assistant/, from the command palette, or the Felix bubble in the panel.
| At a glance | |
|---|---|
| You need | A Falix account with at least one server |
| Plan | All plans — free uses the fast model; premium adds a stronger model and image attachments |
| Time | Ten minutes to get comfortable |
The assistant works on one server at a time. Pick the server from the dropdown, and from then on the conversation is about that server — the assistant can see and act only within it. Switch servers (or start a new conversation) any time.
Once you've picked a server, the assistant can read the things it needs to help:
What it can't see is your secrets. Sensitive values — passwords, tokens, API keys, IPs, emails — are stripped out before anything is sent to the model, so the assistant works without ever handling your credentials in the clear.
🎯 Good to know: The first time you use it, you agree to a short data-sharing notice — it explains that the conversation is AI-powered, that messages are stored, and that server data is shared to answer you. You can revoke that consent later; the assistant stops there without it.
This is where it goes beyond a chatbot. The assistant has a toolbox that spans most of the panel — it can edit and write files, send console commands, take power actions (start / restart / stop), and change properties, settings, schedules, network and firewall rules, players, backups, Git, plugins, and mods. Ask it to "fix the memory error in my config" or "install EssentialsX and set spawn protection," and it works out the steps.
Two limits keep this safe:
A trust setting decides how much the assistant does on its own:
| Level | Behaviour |
|---|---|
| Cautious | Approve every action before it runs |
| Balanced (default) | Reads run freely; changes ask for approval |
| Autonomous | Changes run without prompting — for when you want it to just get on with it |
One rule never bends: destructive actions always ask for approval, even in Autonomous mode. Deleting or wiping something is never done unattended — a safeguard against a malicious instruction sneaking in through a log or a file the assistant reads. Sensitive files like .env also always require a manual OK.
⚠️ Heads up: Autonomous mode is powerful and fast, but it's letting the assistant change your server without a checkpoint. Use Balanced while you're learning what it does, and keep a backup before you turn it loose on anything important.
The assistant itself is on every plan. The differences:
There are fair-use token limits per day and week so the shared service stays fast for everyone; the input box shows when you're near them.
/ai-assistant/, with your conversation history.Conversations are saved per server, so you can pick one back up; delete or close them from the list.