> ## Documentation Index
> Fetch the complete documentation index at: https://docs.firebolt.io/llms.txt
> Use this file to discover all available pages before exploring further.

<AgentInstructions>

## Submitting Feedback

If you encounter incorrect, outdated, or confusing documentation on this page, submit feedback:

POST https://docs.firebolt.io/feedback

```json
{
  "path": "/guides/integrations/mcp",
  "feedback": "Description of the issue"
}
```

Only submit feedback when you have something specific and actionable to report.

</AgentInstructions>

> Use Firebolt MCP Server to enable AI-powered workflows with LLMs.

# MCP Server

Firebolt MCP Server is a lightweight service that enables **large language models (LLMs)** like Claude, GitHub Copilot Chat, and Cursor to connect to Firebolt in a secure, context-aware way. It acts as a bridge between your data warehouse and AI assistants, allowing them to:

* Understand Firebolt-specific SQL
* Query your databases with context-aware prompts
* Access technical documentation, metadata, and live data

It's designed for developers, analysts, and teams who want to integrate Firebolt into AI-driven workflows or copilots with minimal setup.

<Info>
  Full setup instructions, environment variables, and integration guides are available on the [Firebolt MCP GitHub repository](https://github.com/firebolt-db/mcp-server).
</Info>

## When to Use MCP Server

Use Firebolt MCP Server if you want to:

* Enable LLMs to write and execute SQL against your Firebolt environment
* Automate documentation lookups and metadata extraction
* Create advanced AI agents that can explore, troubleshoot, and analyze Firebolt data

Typical use cases include:

* Querying Firebolt using natural language
* Custom copilots in VSCode or Cursor with deep Firebolt context
* AI workflows that require real-time access to Firebolt SQL features

## Getting Started

MCP Server is available as both a binary and a Docker container. To run it, you’ll need:

* A [Firebolt service account](/guides/managing-your-organization/service-accounts) (Client ID and Secret)
* Either Docker or a supported OS for running Go binaries
* An LLM client that supports [Model Context Protocol (MCP)](https://modelcontextprotocol.io)