MCP Agents Overview
MCP agents connect your Vectorize pipelines directly to AI desktop tools like Claude Desktop, Claude Code, Cursor, and other MCP-compatible applications. Access your data and functions where you already work - in your IDE, writing tools, and AI assistants.
What is an MCP Agent?
An MCP agent integrates your Vectorize pipelines into existing AI tools through the Model Context Protocol. Your data becomes available as native tools within applications you already use for development and writing.
When to Use MCP Agents
Choose MCP agents when you want to:
- Enhance AI coding assistants with your codebase knowledge
- Add domain expertise to Claude Desktop or similar tools
- Avoid context switching by keeping data access in your workflow
- Provide specialized functions beyond simple retrieval
Key Concepts
MCP Functions
Functions are the building blocks of MCP agents. Each function:
- Performs a specific retrieval or action
- Can have parameters for filtering
- Maps to tools in your AI assistant
Example functions:
search-docs
- Search technical documentationfind-customer-data
- Query customer records with filtersget-code-examples
- Retrieve relevant code snippets
Connection Methods
Vectorize provides two ways to connect MCP agents:
-
Remote MCP (Recommended) - Vectorize's hosted server
- No infrastructure to manage
- Scales automatically
- Instant updates when you change agent configuration
-
Local MCP Server - Run your own server
- For development environments
- Air-gapped or high-security requirements
- Full control over the connection
Most users choose Remote MCP for its simplicity and reliability.
Next Steps
- Create your first MCP agent - Step-by-step setup guide
- Configure MCP functions - Design effective tools
- Browse integrations - Connect to your favorite tools