MCP Server

📢 Note: The API is currently in Beta.

Vectorize provides a Model Context Protocol (MCP) server implementation that integrates with AI assistants like Claude to enable Vector Retrieval and Deep Research capabilities on your own data.

What is the Model Context Protocol?

The Model Context Protocol (MCP) is a standard for connecting AI assistants to external tools and data sources. It allows AI models to access and utilize specialized capabilities beyond their training data, such as retrieving information from vector databases or processing documents.

Getting started

To configure the MCP server with Claude or any MCP client, you only need to copy the following configuration:

{
  "mcpServers": {
    "vectorize": {
      "command": "npx",
      "args": ["-y", "@vectorize-io/vectorize-mcp-server@latest"],
      "env": {
        "VECTORIZE_ORG_ID": "your-org-id",
        "VECTORIZE_TOKEN": "your-token",
        "VECTORIZE_PIPELINE_ID": "your-pipeline-id"
      }
    }
  }
}

You can find all the information on the "Connect" tab in the pipeline dashboard.

Available Tools

The Vectorize MCP server provides three main tools:

1. Retrieve Documents

Perform vector search and retrieve documents from a Vectorize pipeline:

{
  "name": "retrieve",
  "arguments": {
    "question": "Financial health of the company",
    "k": 5
  }
}

This tool uses the Retrieval API to search for relevant documents in your vector database.

2. Text Extraction and Chunking

Extract text from a document and chunk it into Markdown format:

{
  "name": "extract",
  "arguments": {
    "base64document": "base64-encoded-document",
    "contentType": "application/pdf"
  }
}

This tool uses the Extraction API to process documents and convert them to structured text.

3. Deep Research

Generate a Private Deep Research from your pipeline:

{
  "name": "deep-research",
  "arguments": {
    "query": "Generate a financial status report about the company",
    "webSearch": true
  }
}

This tool uses the Deep Research API to create comprehensive research reports based on your data.

Example Usage with Claude

Here's an example of how to use the Vectorize MCP server with Claude:

1. Configure the MCP server as shown above

2. In your conversation with Claude, you can ask it to use the Vectorize tools:

Search information about our company's financial performance in Q2 2024 in Vectorize.

Claude will then use the MCP server to retrieve the relevant information and provide a response based on the retrieved data.

Next Steps

• Learn more about Vectorize API
• Explore the Retrieval API for vector search
• Discover the Extraction API for document processing
• Try the Deep Research API for comprehensive research generation
• Questions? Join our Discord community for support and discussions. You can also open an issue on the official GitHub repository.