AI Connector (MCP)

Servycore speaks the Model Context Protocol (MCP), the standard AI assistants use to work with other tools. Connect your assistant once and ask it, in plain words:

"Who got reported the most this week, and what did they say in chat?"
"Give Alex VIP for 30 days on every server."
"Why did this player not get his Discord reward?" (it reads the linking command history)
"Answer the open support tickets about kits, show me each reply first."
"Are there plugin updates on EU Main?"
"Write a blog post announcing Thursday's wipe, as a draft."

The full list of what it can do is on the AI Connector page, read from the server itself.

Connect

Open your project and click AI Connector (MCP) in the menu: the page walks you through it for each assistant, with the commands and files to copy.

The address of the server is https://servycore.io/mcp.

Claude (claude.ai, desktop, mobile): Customize, Connectors, Add custom connector, paste the address, then Connect. On a Team or Enterprise plan, an owner adds it first in Organization settings, Connectors.
ChatGPT: Settings, Apps & Connectors, turn on Developer mode under Advanced settings, then Create a connector with the address and OAuth.
Claude Code: claude mcp add --transport http servycore https://servycore.io/mcp, then /mcp to sign in.
Cursor: add {"mcpServers": {"servycore": {"url": "https://servycore.io/mcp"}}} to mcp.json, then Connect.
VS Code: add {"servers": {"servycore": {"type": "http", "url": "https://servycore.io/mcp"}}} to .vscode/mcp.json.

Your assistant opens Servycore: sign in, choose the project it will work in and tick Read only if it should only look. No key to copy.

Assistants that cannot sign in can send an API token instead (Authorization: Bearer …): create it on API Tokens with a Project (a token for every project is refused by the connector).

What keeps you safe

One project per connection. A second project is a second connection.
Your rights, never more. The assistant has exactly what your team role allows: a moderator's assistant moderates and runs only the quick console commands of its role; it never runs free commands, installs plugins or writes blog posts.
It asks first. Acts your players see (messages, kicks, bans, console commands, restarts, groups, ticket replies, published blog posts) are described as such to the assistant, which repeats what it is about to do and waits for your yes. Blog posts are saved as drafts unless you ask to publish them.
Logged. Its acts on servers and players are in the Control Center activity log under your name; kicks, bans, mutes, notes, messages, report handling and automation switches are marked as coming from the assistant and listed on your AI Connector (MCP) account page.
No secrets. RCON passwords and host keys never go through the assistant, and webhooks are masked. It can point a notification card to a new webhook only after you confirm.

What it deliberately cannot do

Wipes are read only: no schedule change and no wipe now. No leaderboard events, no file editing, no server credentials, no billing, team members, project deletion, page builder, themes or store settings. Those stay in the panel; the assistant gives you the link to the right page.

See and disconnect your assistants

AI Connector (MCP) in your account menu lists every connected assistant with its project and access, what your assistants recently did on your servers, and the API tokens you made for one project. Disconnect stops an assistant at once; remove Servycore from the assistant too. The AI Connector (MCP) page of a project shows the assistants connected to that project.

The REST API

Most of what the connector does is also in the REST API for your own scripts and bots.