read_clipboard Tool: Parameters and Formats

read_clipboard reads the current clipboard content with automatic format detection. It is the only tool mcp-clip exposes.

Example usage in Claude:

“Read my clipboard and analyze the content”

Parameters

All parameters are optional.

| Parameter | Description | |---|---| | format | Format to return clipboard content in: text, base64, file, or auto (default) | | output_path | Save the clipboard content directly to the specified file path, instead of returning it inline | | chunk_size | For large content, split into chunks of this size in bytes. Default: 25000 | | chunk_index | Return a specific chunk index (0-based) for large content |

Supported formats

• Plain text
• Images (PNG, JPEG, GIF, WebP, BMP)
• Binary data (base64 encoded)

Large content handling

• Content over chunk_size (default 25000 bytes) is returned as numbered chunks, or as a pointer to the mcp-clip://current resource — request a chunk with chunk_index, or read the resource directly
• Pass format: "file" or output_path to write the content to disk instead; on Windows this path uses PowerShell to extract the clipboard, which is how images and file lists get saved
• File paths are returned in the tool result for external access

The mcp-clip://current resource

The server also registers a mcp-clip://current resource that always reflects the current clipboard content:

{
  "uri": "mcp-clip://current"
}

A client can read this resource directly instead of calling read_clipboard again. The server tracks clipboard changes internally as they happen, but does not yet push notifications/resources/updated events — reading the resource or the tool is still poll-based.

See also

• Getting Started — install mcp-clip and add it to your MCP client
• WSL2 setup — when output_path extraction goes through PowerShell