How to Set Up a MySQL MCP Server With Gemini CLI

It was a Tuesday and someone on the ops team needed a quick number: how many distinct products did we actually sell in March, not counting refunds or test orders? A fifteen-minute Slack thread ensued because the engineer who knew the schema was in a different timezone and the one who was available wasn't sure which join to use. I pulled up Gemini CLI, pointed it at the database through QueryBear, and had the answer in ninety seconds. That Slack thread is why I set this up.

Part of the MySQL MCP Server guide.

What you'll need

• A MySQL database (read replica recommended)
• A QueryBear account
• Gemini CLI installed (npm i -g @google/gemini-cli or via Homebrew)

Why route through a gateway

Passing a raw mysql:// URL to an agent means the agent is working with whatever privileges that user has. A read-only MySQL user helps, but a gateway gives you more: parser-level DML rejection, a table allowlist, column masking, per-query timeouts, and an audit log. The detailed case is in the MySQL MCP Server guide.

Connect QueryBear to your MySQL

1. Open the QueryBear dashboard and click "New connection."
2. Choose "MySQL" and paste your connection URL. Use a replica and a user with SELECT privilege only.
3. Under "Allowed tables," add the tables that are relevant. Skip anything you wouldn't want quoted in a Slack message.
4. Under "Blocked columns," add things like password_hash, api_key, credit_card_last4. Save.

QueryBear will sit between Gemini and your database from here on.

Install the QueryBear MCP server in Gemini CLI

Edit ~/.gemini/settings.json (create it if it doesn't exist):

{
  "mcpServers": {
    "querybear": {
      "httpUrl": "https://mcp.querybear.com/mcp"
    }
  }
}

That's the full config. Gemini CLI reads settings.json on each invocation, so there's nothing to restart.

Ask your first question

I asked: "How many distinct products did we sell in March?"

Gemini called get_schema, looked at the tables, then wrote this:

SELECT COUNT(DISTINCT oi.product_id) AS distinct_products
FROM order_items oi
JOIN orders o ON o.id = oi.order_id
WHERE o.created_at BETWEEN '2026-03-01' AND '2026-03-31';

QueryBear validated the query (allowlisted tables, no DML, columns not blocked), ran it against the replica, and returned a single row. Gemini printed the count inline in the terminal with a short explanation of what the join was doing.

One thing worth knowing about MySQL: the BETWEEN bounds are inclusive on both ends, so '2026-03-31' captures records up to midnight on March 31, not through the end of the day. If your created_at stores time components, use < '2026-04-01' for the upper bound. Gemini didn't hit this edge case in my schema because everything was date-only, but it's worth catching early.

A Gemini-specific tip

Gemini CLI has a large context window, which means you can pipe its output into other tools without worrying about truncation. If you want a CSV instead of a markdown table, try something like:

gemini "Run a QueryBear query: how many distinct products did we sell in March? Output as CSV." | tee march-products.csv

The result goes to the file and to your terminal at the same time. Useful for dropping quick exports into a spreadsheet without opening anything else.

Using a different tool?

• Claude Code
• Claude Cowork
• Cursor
• Windsurf
• ChatGPT
• Codex

Querying a different database?

• Postgres MCP Server
• SQLite MCP Server