# MCP Servers

> You're reading the Mindset v3 documentation. v3 is no longer sold; this is reference for existing customers. Mindset 4 docs are at /docs.

What’s an MCP server?

**MCP** stands for **Model Context Protocol** - an open standard that lets AI models (like large language models) connect to real-time data, external tools, and other systems in a consistent, scalable way.

Think of it like **USB-C for AI**: once a tool “speaks MCP,” any compatible model can use it without custom wiring or special code.

Instead of brittle, one-off integrations, MCP creates a **universal layer** between your model and the outside world - APIs, databases, calculators, search engines, or internal tools. This means you can **build once, reuse everywhere**, and easily swap components without breaking things.

## Without MCP

- Every agent needs a custom integration for each tool.
- You spend time reinventing the wheel for every connection.
- Communication between the agent and tools can be messy, error-prone, and inconsistent.

## With MCP

- There’s a standard “language” for exchanging data and instructions.
- You can swap tools in and out without rewriting the agent.
- Agents can understand new tools faster because their capabilities are described in a consistent way.

## Types of MCP Servers

Mindset AI supports multiple types of MCP servers to fit your specific needs and use cases:

### 1. Native MCP Server

The Native MCP Server provides a standardized way to access Mindset AI’s built-in workflows as tools for your agents:

- Each workflow in your Mindset environment becomes available as a tool
- Agents receive a list of available tools with descriptions of when to use each one
- Simplifies agent-to-workflow connections without requiring custom code

### 2. Your MCP Servers

#### a. Tools MCP Server

Host your own MCP server to expose your internal systems and APIs as tools for your agents:

- Maintain full control over which tools and capabilities you expose
- Integrate your existing APIs and services 
- Enable agents to perform useful actions within your specific environment
- Access your proprietary systems securely while keeping sensitive implementation details private

[Tools MCP Server integration](https://docs4.mindset.ai/v3/deploy/mcp/client-tool-mcp)

#### b. RAG MCP Server

For organizations with specific requirements around knowledge management and data sovereignty:

- Host your own Retrieval-Augmented Generation (RAG) system while maintaining Mindset AI’s agent capabilities
- Keep sensitive documents and knowledge contexts on your infrastructure
- Maintain complete control over your knowledge sources and how they're accessed
- Provide source references via URLs to maintain citation capabilities

This option is ideal for organizations with existing RAG infrastructure or strict compliance requirements around document storage.

[RAG MCP Server integration](https://docs4.mindset.ai/v3/deploy/mcp/rag-mcpserver-integration)

## Why Use MCP for Agents

### Faster Integrations

Agents can connect to new tools without writing custom code each time because MCP provides a common structure - like plugging devices into a USB port.

### More Reliable Conversations

MCP clearly defines how requests and responses are structured, reducing errors and misunderstandings between agents and tools.

### Tool Interchangeability

Replace a data source or service without breaking the agent’s functionality - minimizing downtime and making it easy to test new options.

### Easier Scaling

Multiple agents can share the same MCP connections, reducing repeated integration work as you scale.

### Future-Readiness

MCP is becoming an industry standard. Adopting it now prepares you for a growing range of compatible tools and agents.

## The Big Picture

MCP is more than a way to connect a model to a calculator or a search API.\
It’s a **flexible, reusable, and secure foundation** for intelligent applications.\
As your systems grow, MCP keeps integrations **clean, maintainable, and ready for future expansion**.

Setting up your MCP servers in the AMS