MCP server¶

An MCP server (in mcp/) exposes the suite to AI agents such as Claude Desktop. It's a thin Python wrapper that shells out to the built folder_* executables over stdio — the C++ tools stay dependency-free; only this wrapper needs Python.

Tools¶

Tool	Wraps	Notes
`compare`	`folder_parity`	report differences; `use_hash`, `ignore_system`
`sync`	`folder_parity --sync`	`sync_to`, `prefer`
`find`	`folder_find`	`name`, `contains`, size/age filters
`view`	`folder_view`	text or `as_hex`
`clean`	`folder_clean`	dry-run unless `delete=True`
`copy`	`folder_copy`	`overwrite`
`move`	`folder_move`	move / rename
`delete`	`folder_delete`	dry-run unless `force=True`

Destructive operations are dry-run by default; sync/copy/move only ever add or overwrite.

Setup¶

Build the CLI tools first, then:

cd mcp
pip install -r requirements.txt
python server.py            # stdio transport, for an MCP client

The server finds the executables via $FOLDER_TOOLS_DIR, else the repo root.

Claude Desktop¶

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "folder-tools": {
      "command": "python",
      "args": ["D:/Github-Sina/folder_parity/mcp/server.py"],
      "env": { "FOLDER_TOOLS_DIR": "D:/Github-Sina/folder_parity" }
    }
  }
}

Then: "compare J:\Photos and I:\Photos by hash", "find *.log over 1 MB under D:\logs", "clean OS junk in this folder (dry run)".