MCP Integration

The MCP (Model Context Protocol) integration lets you easily connect your preferred LLM providers to your SmartBucket. This enables secure, unified access to your data and language models, making it easy to configure, authenticate, and retrieve content from SmartBuckets across different LLM environments.

Prerequisites

1. Raindrop API Key
   • The Raindrop SDK uses API keys for authentication.
   • You can obtain your Raindrop API key by logging into your Raindrop Account and creating a new API key in Settings -> API Keys.
2. An LLM Provider
   • You can use any LLM provider that supports the MCP protocol.
   • Below we will cover the general configuration for most MCP integrations and two examples of desktop LLM providers’ MCP integrations (Claude and Cursor).

General Configuration for your MCP Integration

Note

Please note that the configuration below is a general configuration that can be used for most LLM providers. For LLM providers that support Streamable HTTP, you can use the https://mcp.raindrop.run/mcp endpoint instead of the https://mcp.raindrop.run/sse endpoint.

MCP Configuration File

{
    "mcpServers": {
        "liquidmetal": {
            "command": "npx",
            "args": [
                "mcp-remote",
                "https://mcp.raindrop.run/sse",
                "--header",
                "Authorization: Bearer ${RAINDROP_API_KEY}"
            ],
            "env": {
                "RAINDROP_API_KEY": "<ADD_YOUR_RAINDROP_API_KEY_HERE>"
            }
        }
    }
}

Tools

After configuring your MCP integration with SmartBuckets, you now have access to the following tools:

• document-search
• document-query
• chunk-search

Tip

For all prompts, please specify the smartbucket-name in your query.

For example, if your smartbucket name is history-smartbucket, you can specify it in the query like this: Find all content related to the topic of knights in my SmartBucket called "history-smartbucket".

Get your smartbucket name by going to your manifest file or running raindrop build find -o json in your smartbucket app terminal.

Document Search

Use the document-search tool to search for documents in your SmartBucket. The input is a natural language query and the output is a list of documents that match the query.

Example: Find all PDF documents in my SmartBucket called "history-smartbucket".

Document Query

Use the document-query tool to firectly ask questions about a specific document in your SmartBucket. The input is a natural language query with an output and the output is a natural language response to the query.

Example: What is the name of the person in document medieval-times.pdf? In my SmartBucket called "history-smartbucket".

Chunk Search

Use the chunk-search tool to search for content in all your documents in your SmartBucket. The input is a natural language query and the output is a list of content that matches the query.

Example: Find all content related to the topic of knights in my SmartBucket called "history-smartbucket".

Claude MCP Integration

For the Claude MCP integration, add the following to your claude_desktop_config.json file:

Tip

To go to your claude_desktop_config.json file, go to Claude Desktop App -> Settings -> Developer -> Edit Config -> File: claude_desktop_config.json

claude_desktop_config.json

{
    "mcpServers": {
        "liquidmetal": {
            "command": "npx",
            "args": [
                "mcp-remote",
                "https://mcp.raindrop.run/sse",
                "--header",
                "Authorization: Bearer ${RAINDROP_API_KEY}"
            ],
            "env": {
                "RAINDROP_API_KEY": "<ADD_YOUR_RAINDROP_API_KEY_HERE>"
            }
        }
    }
}

SmartBucket Tools

After adding to the configuration file, you can see your tools in your Claude Desktop App.

For more information about Claude’s MCP integration go to Claude’s MCP integration docs.

Cursor MCP Integration

To get started with the Cursor MCP integration, add the following to your .cursor/mcp.json file:

.cursor/mcp.json

{
  "mcpServers": {
    "liquidmetal": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.raindrop.run/sse",
        "--header",
        "Authorization:${AUTH_HEADER}"
      ],
      "env": {
        "AUTH_HEADER": " Bearer <ADD_YOUR_RAINDROP_API_KEY_HERE>"
      }
    }
  }
}

For more information about Cursor’s MCP integration go to Cursor’s MCP integration docs.