AI Assistants (MCP)
Connect Claude, Cursor and other AI assistants to your Supascale install.
Supascale supports the Model Context Protocol (MCP), an open standard for connecting AI assistants to external systems. Once set up, you can manage your projects by asking:
"Which of my projects has a backup older than a week?"
"Restart the staging project — it stopped responding."
"Back up production before I run this migration."
Works with Claude, Claude Code, Cursor, VS Code, Windsurf, and any other MCP-capable client.
How it works
A small connector runs on your own machine — your AI assistant launches it — and talks to your Supascale server over its REST API.
This means your Supascale server does not need to be reachable from the internet. A LAN address, a VPN hostname or a tunnel all work fine.
The connector holds no permissions of its own. It carries an API key, and your server decides on every single call whether that key may do what was asked.
What an assistant can do
| Area | Actions |
|---|---|
| Projects | List, inspect, check status, read logs, start, stop, restart, create |
| Backups | List, check status, create |
| Tasks | List scheduled tasks, inspect one |
| System | Host info, CPU/memory/disk usage, check for updates |
What it deliberately cannot do
Deleting a project and restoring a backup are not available through MCP. Not disabled — absent. Both are irreversible, and an assistant deciding to call one is a bad enough outcome that we did not build the tools at all. There is no setting that turns them on.
Cloud storage credentials are also out of reach.
Quick setup
- Settings → MCP → turn on Enable MCP.
- Settings → API Keys → create a key with Allow MCP access enabled.
- Paste the config snippet shown on the MCP settings page into your assistant.
Full walkthrough: Setting up MCP.
Next steps
- Setting up MCP — per-client configuration
- Available tools — every tool and the permission it needs
- MCP security — how access is decided, and what is logged