ClickHouse MCP Server

🔍 What is MCP?

MCP (Model Context Protocol) is a lightweight protocol that acts as a communication bridge between large language model (LLM) clients and servers (MCP Server). With Squel MCP Servers, It enables natural language queries to be translated into SQL in real time — securely and locally — without requiring users to write any code or expose sensitive information.

In short, MCP makes your AI/MCP Client speak with external tools (your database) using Sequel's MCP servers.

🛡️ Is it secure? Will my data be exposed to the internet?

Your data never leaves your machine.

Sequel's MCP servers are designed to run locally on your machine or inside your private infrastructure. There’s no external syncing, telemetry, or cloud relay involved. The communication between the AI client (e.g., Cursor, Claude) and your database happens entirely over local HTTP/SSE/IO transports.

This makes it ideal for sensitive environments like finance, healthcare, and enterprise systems where data compliance is critical.

🧠 How do MCP clients like Cursor or Claude use this?

Once the MCP server is running, it exposes a local endpoint that adheres to the Model Context Protocol standard. MCP-compatible clients (like Cursor or Claude) are pre-configured to discover this endpoint and send context-aware natural language queries.

The flow looks like this:

1. You install the MCP server for your database.
2. You launch the MCP-compatible client (e.g., Cursor).
3. The client detects the running MCP server.
4. You type a natural question (e.g., “What was our revenue last quarter?”).
5. The server translates it to SQL, runs it on database, and returns results.

🛠️ What are the supported databases?

MCP currently supports the following databases out of the box:

• PostgreSQL — postgresql
• MySQL — mysql
• ClickHouse — clickhouse

To install the server for a specific database, run:

npx -y @sequelsh/mcp@latest install -d <database>

💻 What MCP clients are supported?

MCP integrates smoothly with the following clients:

• 🧠 Claude — for conversational, context-aware AI queries.
• 💻 Cursor — an AI-powered IDE that interfaces directly with your database via MCP.
• 🌊 WindSurf — a UI-based analytics tool with LLM assistance.

Install support for a client using:

npx -y @sequelsh/mcp@latest install -c <client>

🔁 What transport types are available, and when should I use each?

The MCP server supports the following transport protocols:

• SSE (Server-Sent Events) — default, best for streaming query results to AI clients.
• IO (Socket-like) — recommended when working with interactive or bi-directional clients.

If you’re unsure, start with the default IO.

🐳 Can I run MCP inside Docker or deploy it to a remote server?

Right now, MCP is optimized for local use only — especially for use cases where developers interact with their local or VPN-bound databases.

That said, Docker support and deployment templates are in the roadmap. If you're building something custom, you could wrap the CLI inside a container manually, but official support isn’t there yet.

🧪 How do I test if the MCP server is running?

Once you have installed the MCP servers, you can check your MCP client to see if the server is running.

❌ How do I uninstall or reset an MCP server?

Currently, there isn’t a dedicated uninstall command. However, you can reset your environment by doing the following:

Open the configuration file for your MCP client and remove the entry of Sequel's MCP server

🤝 How can I contribute or request new features?

We're open to contributions! Visit the official GitHub repo:

🔗 https://github.com/haxzie/sequel-mcp

You can:

• Submit pull requests
• Open feature requests
• Report bugs
• Request support for a new database or MCP client