MCP Tools Integration

Extend your chatbot's capabilities with Model Context Protocol tools

What are MCP Tools?

Model Context Protocol (MCP) tools allow you to extend your chatbot's capabilities beyond simple question-answering. These are custom functions that the LLM can call to perform specific actions, access external APIs, or execute business logic.

ragistry's MCP implementation enables the AI to intelligently decide when to use tools, chain multiple tools together, and integrate their outputs into natural language responses.

Built-in Tools

ragistry comes with several pre-built MCP tools for common use cases:

Database Query Tool

Execute SQL queries against connected databases to fetch real-time data. Supports parameterized queries for security and automatic result formatting.

Search Enhancement Tool

Perform advanced searches with filters, date ranges, and custom relevance scoring. Useful for narrowing down large knowledge bases.

Analytics Tool

Retrieve usage statistics, conversation metrics, and performance data. Enables the chatbot to answer questions about system usage.

External API Tool

Make HTTP requests to external services with authentication, rate limiting, and error handling. Integrate third-party data into responses.

Creating Custom Tools

Build your own MCP tools to handle specific business logic:

from models.mcp_tools import MCPTool
from typing import Dict, Any

class CustomTool(MCPTool):
    name = "custom_tool"
    description = "Performs a custom action"
    
    async def execute(self, params: Dict[str, Any]):
        # Your custom logic here
        return {"result": "success"}

Tool Management

Configure and manage your MCP tools through the dashboard:

  • Enable or disable tools per organization
  • Set tool parameters and configuration
  • Monitor tool usage and performance
  • View tool execution logs and errors
  • Test tools with sample inputs

Security & Permissions

MCP tools execute server-side with appropriate security measures:

Authentication

Tools inherit user permissions and can be restricted based on roles

Input Validation

All parameters are validated before execution to prevent injection attacks

Rate Limiting

Prevent abuse with configurable rate limits per tool and organization

Audit Logging

Complete audit trail of tool executions with timestamps and results

Use Cases

E-commerce

Check inventory levels, retrieve product details, process orders, track shipments

Customer Support

Look up customer accounts, check ticket status, escalate issues, send notifications

Internal Tools

Query internal databases, generate reports, schedule tasks, update records

MCP Service Architecture

Tools are managed by the mcp_service.py and mcp_tool_service.py modules, with support for both synchronous and asynchronous execution. The system includes circuit breakers and fallback mechanisms for reliability.