Plesk add-on

The Plesk MCP server connects Claude to the Plesk panel on your own hosting server. Claude can look up domains, clients, DNS records, databases, FTP users, extensions and domain logs, and read files under a domain's document root. In read-write mode it can also create, change and delete those objects and run commands on the server.

The Plesk MCP server is an add-on, not a standalone server. It runs inside mcp-core, the self-hosted MCP server that you install with Docker. mcp-core talks to the Plesk REST API (v2) with an API key that stays in the .env file on your own machine.

What can Claude look up through the Plesk MCP server?

In read-only mode the Plesk MCP server gives Claude 25 tools that read your panel and change nothing. They cover:

• Server info (panel version, hostname, platform) and the server's IP addresses
• Installed extensions with version and active state
• Clients: details, the domains they own and their resource usage
• Domains: details, owner, status, and Plesk's built-in DNS resolution check
• A search through a domain's web, mail and PHP logs by pattern and time range
• DNS records, databases, database users, database servers and FTP users
• Domain files: list a directory, get file metadata, download a file's contents
• The list of Plesk CLI utilities and the parameter reference for each one

That makes prompts possible like "which client owns this domain and is it suspended?", "check the DNS of example.com and list its MX records" or "search the PHP log of this site for fatal errors since this morning". Claude chooses the tool and the parameters.

What does read-write mode on the Plesk MCP server allow?

Read-write mode on the Plesk MCP server adds 26 tools that change your server. Claude can create, update, suspend, activate and delete clients; create, update and delete domains and set their status; manage FTP users, databases, database users and DNS records; and install, uninstall, enable or disable extensions.

Two tools go further. plesk_exec_domain_command runs any shell command as a domain's hosting user, by default in its document root. plesk_call_cli_command runs a Plesk CLI utility with server-wide administrative scope, not limited to one domain. Both are marked DANGEROUS in their tool description.

How risky is write mode on a hosting server?

Very. A read-write seat on the Plesk MCP server can take sites offline or destroy data in one call. Deleting a client removes everything it owns: domains, mailboxes and databases. Deleting a domain removes its hosting, mail, databases and files. Deleting a database removes its data. None of these can be undone.

mcp-core asks for no second confirmation. When Claude calls a write tool of the Plesk MCP server, the request goes straight to the Plesk API. The only safeguards in the tool are the mode gate and loud warnings in the tool descriptions ("PERMANENTLY delete", "This cannot be undone", "DANGEROUS"), so Claude sees the impact before it calls a tool. Treat a read-write seat like handing someone your admin login.

Our advice: give read-write only to a seat that you use yourself, for a task you asked for explicitly, and keep every other seat read-only. Make sure a recent backup exists before you let Claude change anything.

What is blocked in every mode?

Some Plesk endpoints are left out of the Plesk MCP server on purpose, also in read-write mode:

• Writing to domain files. Upload, delete, copy, move, mkdir, append and chmod are not implemented. Claude can read files, never write them through the file tools.
• API key management. The tool logs in with a key; it cannot create or revoke keys, including its own.
• Server initialisation and the Plesk license endpoints.

Note that the shell command tool can still change files inside a domain. The file block applies to the file tools, not to commands you allow Claude to run.

How do you set up the Plesk MCP server?

You need a running mcp-core install, the Plesk add-on on the same license and a Plesk API key. Then:

1. Buy the Plesk add-on. It is added to your existing license; there is no new download.
2. Add the two variables below to the .env next to your docker-compose.yml.
3. Run ./mcp-update.sh, then docker exec mcp-core npm run refresh-license to pick up the entitlement right away.

PLESK_BASE_URL=https://panel.example.com:8443
PLESK_API_KEY=your-plesk-api-key

PLESK_BASE_URL is the bare panel address you type in a browser. mcp-core adds /api/v2 itself. The key is sent in the X-API-Key header; your admin password never goes into the container. Without both variables, no Plesk tools are registered. The Plesk MCP setup guide covers every step and the error messages.

Can each seat get its own Plesk access?

Yes. Each seat on the Plesk MCP server is one person or device with its own token, its own subset of tools and its own read-only or read-write mode. You choose the mode when you add the seat in My Account → MCP Self-hosted → Seats. On a read-only seat the write tools are not registered at all, so Claude cannot call them. How the license and the mode work together is explained in entitlements vs. tool modes.

MCP Self-hosted includes three seats. A support colleague who only needs to read logs and DNS gets a read-only seat; add an extra seat when more people need access.

Who is the Plesk MCP server for?

It is built for agencies, hosting resellers and developers who run client sites on their own Plesk server and already work with Claude. Typical read-only work: find which client owns a domain, check DNS after a migration, trace an error in the logs, compare database users across subscriptions. With write access: add a DNS record, create a database and user for a new site, suspend a client.

It combines well with the WordPress MCP server: Claude checks the hosting side in Plesk and the content side in WordPress in the same conversation.