---
updatedAt: 2026-03-25T01:06:51.000Z
agentTools:
  projectIndex: https://docs.getmontecarlo.com/llms.txt
---

# Monte Carlo MCP Server Technical Overview

The Monte Carlo <Anchor label="Model Context Protocol (MCP) Server" target="_blank" href="https://docs.getmontecarlo.com/docs/mcp-server">Model Context Protocol (MCP) Server</Anchor> enables AI applications and agents such as <Anchor label="Cursor" target="_blank" href="https://www.cursor.sh/">Cursor</Anchor>, <Anchor label="Claude" target="_blank" href="https://www.claude.com/download">Claude</Anchor>, and other MCP compatible tools to interact securely with Monte Carlo’s Data + AI Observability Platform.

The Monte Carlo MCP Server provides a secure, scalable, and fully managed environment for AI-assisted data quality management. It connects Monte Carlo’s trusted data quality, lineage, and alerting APIs with AI agents, while maintaining enterprise-grade security.

***

## Architecture Summary

### Core Components

| Component               | Description                                                                            |
| ----------------------- | -------------------------------------------------------------------------------------- |
| **MCP Clients**         | AI tools that initiate requests through the MCP protocol.                              |
| **Integration Gateway** | Authenticates requests, validates tokens, and forwards traffic to the MCP Server.      |
| **MCP Server**          | Hosted in AWS Lambda, exposes Monte Carlo’s capabilities such as alerting and lineage. |
| **GraphQL API**         | Monte Carlo’s backend service responsible for executing and authorizing data queries.  |

***

## Authentication & Authorization

The MCP Server enforces strict authentication and authorization aligned with both the MCP Protocol and Monte Carlo’s security and compliance frameworks.

### OAuth 2.1

* You can use OAuth 2.1 to authenticate clients with the Monte Carlo MCP server.
* Users must authenticate and consent in order to get tokens to call the MCP server.

### API Keys

* Customers authenticate using their standard Monte Carlo credentials.
* The MCP Server never directly reuses customer tokens, preventing *token pass-through* violations.
* API calls are executed under the user’s identity in Monte Carlo’s GraphQL API, preserving existing access controls.

***

## Data Access & Privacy

* **Customer Isolation:** All API calls are scoped to the authenticated user’s tenant and data region.
* **Encryption in Transit:** All communications between clients, gateways, and the MCP Server use HTTPS (TLS 1.2+).
* **Stateless Design:** The MCP Server does not persist or cache customer data — all processing happens in real time.
* **Auditing:** Requests include telemetry metadata for full traceability and compliance review.