Give your AI agent real access to your hosting
One endpoint, one scoped token. Connect any MCP-compatible AI and let it run real work across every service you own.
Claude
ChatGPT
Cursor
VS Code
KiroOne agent, your entire stack
Eleven areas of real control — each granted only when you say so, per token.
File Access & Editing
Browse, read, edit & search files anywhere you grant access.
Database Access
Run SQL and manage MySQL, Postgres & users.
WordPress Management
Plugins, themes, core updates & security scans.
Node.js Management
Installs, PM2 processes & Nginx reverse proxy.
Python Management
Virtualenvs, pip packages & WSGI/ASGI apps.
Docker & Container Control
Shell in, control lifecycle, read stats & backups.
Server & VM Control
SSH commands, systemd services & power state.
Domains & DNS
DNS records, WHOIS, transfers & Cloudflare.
Email Management
Create mailboxes, read, send & route mail.
Customer Conversations
Read & reply to customers across every channel.
Billing & Support
Tickets, invoices, team & plan upgrades.
Order Assistant
always onBrowse the catalog, price a cart & place orders — no grant needed.
From your chat to your server, in one hop
Broodle MCP is the secure bridge between your AI and your hosting. Here's a single request — “restart my app and check the logs” — travelling end to end.
You ask in plain English, inside the tool you already use.
appcontainer.restart_appYour scoped token authorizes it; MCP maps it to the right tool call.
It runs on your real container, then reports the result back.
Connected in under two minutes
One endpoint for every client. Pick yours, copy the config, drop in a token — done.
your client
https://broodle.host/api/mcpOpen Settings → Developer → Edit Config, paste this in, then restart Claude Desktop.
{
"mcpServers": {
"broodle": {
"type": "http",
"url": "https://broodle.host/api/mcp",
"headers": { "Authorization": "Bearer YOUR_TOKEN_HERE" }
}
}
}Replace YOUR_TOKEN_HERE with a token from Console → Tools → MCP Server.
You mint the token. You set the limits.
Scoped per service, tagged by risk, expiring, and fully logged. Nothing runs beyond what this token allows.
Edit files, restart apps, add records, send mail.
Granular, per-token grants
A token only works on the exact services, domains, and account permissions you selected when creating it — never your whole account by default.
Optional IP allowlisting
Restrict a token to specific IP addresses or CIDR ranges so it only works from where you expect — off by default, one toggle to turn on.
Full usage history
Every tool call is logged — which tool, which token, from which IP, whether it succeeded — visible right in your console with filters and search.
Expiring tokens
Tokens expire automatically (1 day up to 90 days, your choice) and can be revoked instantly at any time — no residual access left behind.
Secrets are never exposed
Passwords, API keys, and other sensitive values are automatically redacted before anything is logged — even in the detailed call history.
Connected in under two minutes
- 01
Generate a token
Go to Console → Tools → MCP Server and create a token. Pick exactly which services, domains, and account permissions it can touch, and how far it can go — read-only, read & write, or full access including deletes and orders.
- 02
Add it to your AI client
Paste one URL and your token into Claude, Cursor, VS Code, Claude Code, OpenCode, Kiro, or any MCP-compatible client. One connection covers every hosting type, domain, and account capability you granted — no separate integration per service.
- 03
Just ask
Tell your AI agent what you want in plain English — "restart my app container", "check why my WordPress site is slow", "add an MX record for mail.example.com", "open a support ticket about my invoice". It calls the right tool automatically.
Frequently asked questions
Everything you need to know about connecting AI agents to your Broodle hosting via MCP.
MCP (Model Context Protocol) is an open standard that lets AI assistants like Claude, ChatGPT, and coding agents call real tools — not just chat. Broodle’s MCP server exposes your hosting, domains, and account as a set of tools any MCP-compatible AI client can use, so you can manage your infrastructure by asking your favorite AI assistant instead of clicking through a dashboard.
Any client that supports the MCP standard over HTTP: Claude Desktop, Claude Code, Cursor, VS Code, OpenCode, Kiro CLI, and ChatGPT (via custom connectors on supporting plans), among others. It’s one standard connection — if your AI tool speaks MCP, it works with Broodle.
No. Broodle runs a single, unified MCP server at one URL. One token can be granted access to any mix of App Container apps, cPanel hosting, VPS/LXC servers, Engage helpdesk accounts, domains, and account-level permissions — all through the same connection.
You stay in control of exactly what a token can do. Every tool is scoped to read, write, or danger, grants are per-service and per-domain (never blanket account access unless you choose that), tokens expire automatically, and every single call is logged with full detail. Start with a read-only token for anything you’re unsure about.
Practically anything you could do yourself in the console or cPanel — restart an app, read and fix a config file, add a DNS record, install a WordPress plugin, check disk usage, open a support ticket, reply to a customer on WhatsApp via Engage, or even place a real order. It calls the exact same tools our own console and AI Studio use internally.
Yes. Every token supports an optional IP allowlist — add specific IPs or CIDR ranges and the token will only be accepted from those addresses. It’s off by default so tokens work from anywhere, and you can turn it on (or edit the list) at any time from the token’s settings.
The MCP server itself is free — it’s included with every Broodle hosting service, domain, and account. You only pay for the underlying hosting, domains, or orders your AI agent might place on your behalf, exactly as if you’d done it yourself through the console.
Yes — every tool call is recorded in your account’s usage history: which tool ran, on which service, from which token and IP address, whether it succeeded, and (for troubleshooting) the actual call arguments with secrets automatically redacted. Filter by tool, status, or token right from the console.
Ready to let your AI do the work?
Generate a token, add one URL to your AI client, and start managing your hosting by chat — no extra cost, no separate integrations.