AI agents and Syncplify Server!

Since version 8.10.0, Syncplify Server! comes with ss-mcp, a small separate program that connects an AI agent to the server's management API through the Model Context Protocol (MCP), the standard Claude Desktop, Claude Code, Cursor, Visual Studio Code and most other agents use to reach tools. Once connected, you ask the agent in plain language ("which users have not signed in this year?", "create a read only account for the auditor", "why was 203.0.113.7 banned?") and the agent calls the server for you.

Three things make it safe to point an agent at a production server:

• The agent has exactly the rights of one automation token. Read tools for the permissions the token may view, write tools only where it may edit, and nothing at all about operator accounts, passwords, second factors, backups or tokens. A read only token gives a read only agent.
• Every change is previewed, then confirmed. The agent first receives a server side dry run (what would change, field by field) and a confirmation token; it must show you the preview and call again to apply it. A refused dry run cannot be confirmed.
• Everything is audited. Each call is recorded on the server with the token and the operator who owns it, and revoking the token stops the agent at once.

The adapter administers the server; it never lists folders or reads files. An agent that must move files uses SFTP or FTPS with a user account, like any other client.

What you need

• Syncplify Server! 8.8.0 or newer (8.10.0 for the complete set of tools), Professional or Ultimate edition.
• Automation tokens turned on by a SuperAdmin, under Global Configuration, Automation.
• An automation token for the operator the agent will act for.
• An MCP client: Claude Desktop, Claude Code, Cursor, Visual Studio Code or any other program that starts stdio MCP servers.

The articles in this group:

• Download and install the MCP adapter: where to get ss-mcp, how to verify and install it on Linux and Windows, where it should run.
• Connect your AI agent: creating the token and configuring Claude Desktop, Claude Code, Cursor, Visual Studio Code and other clients, with the token kept out of every file and command line.
• What to ask your AI agent: worked examples of questions, changes and reviews, what the confirm step looks like, and what the agent cannot do.

If something does not work, start with The AI agent does not see a tool, or a call fails with 403 or 412. The reference is the manual's chapter Automation and AI agents, in particular The MCP adapter and Automation tokens.