For startups

The MCP gateway to your database, built for startups.

Five people, competing with fifty. QueryBear hands your database to AI agents so you ship instead of building dashboards.

Start free See how it works

Why startups choose QueryBear

Start free

Move fast, no dashboards

Ask your agent for revenue, signups, or churn. Answer back in Slack.

Automate on-call

Agents poll your data and ping you the moment something is off.

Never break prod

Every query is read-only. Agents can look, never touch.

Built for startups

Agents that run the busywork.

Skip the BI tool. Ask your database in English.

Build this with QueryBear How it works

You ask

"Revenue from new accounts this week vs last week?"

Agent queries QueryBear

querybear.run_query(sql: "SELECT SUM(amount) FROM invoices WHERE created_at >= '2026-05-19' AND new_account = true")

→ this week: $84,200 · last week: $72,150

Agent answers

New-account revenue is up 16.7% week-over-week.

One platform, total control.

Everything you need to give AI agents safe database access — from schema discovery to audit logging.

querybear — Auto-detect schema

$ querybear schema refresh --connection prod-postgres
Connecting to prod-postgres...
Scanning tables...  ████████████████████  100%
✓ Found 24 tables, 186 columns, 12 foreign keys
 
  users           14 columns
  orders          11 columns
  products         9 columns
  subscriptions    8 columns
  invoices         7 columns
  payments        10 columns
  teams            6 columns
  addresses       12 columns
  reviews          5 columns
  ... 15 more tables
 
Detecting relationships...
✓ 12 foreign keys mapped
✓ Schema cached and ready for queries

All local, own your data.

Every query, credential, and result stays on your engineer's machine. Nothing touches the cloud.

Download for macOS

Native Desktop App

A macOS app. No browser, no cloud, no middleman.

Local Network Only

Queries go from your machine to your database. Nothing in between.

Zero Data Exposure

Credentials, schema, and results never touch an external server.

Works with your database.

Connect Postgres, MySQL, or SQLite in seconds — including every managed flavor built on them. Same safety layers, same audit trail, zero config.

Core engine

PostgreSQL

Full schema detection. Connect with just your credentials.

Core engine

MySQL

Works out of the box — same safety layers, zero config.

Core engine

SQLite

Lightweight, local, and embedded. Great for prototyping.

Neon

Postgres

Supabase

Postgres

AWS Aurora

Postgres

Amazon RDS

Postgres

Cloud SQL

Postgres

Timescale

Postgres

CockroachDB

Postgres

Railway

Postgres

PlanetScale

MySQL

MariaDB

MySQL

Aurora MySQL

MySQL

TiDB

MySQL

Turso

SQLite

libSQL

SQLite

Fine-grained access control.

Define exactly which columns each role can see. Sensitive data stays hidden from the people who don't need it.

access_policy

Role	id	name	email	phone	address	dob	salary	ssn	bank_account
Admin
Engineer
Analyst
Support
Marketing

Connect your AI agent.

QueryBear exposes an MCP server your AI agents can use to safely query your databases — no extra setup required.

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

Paste into .cursor/mcp.json in your project root.

get_schemarun_query

claude mcp add querybear https://mcp.querybear.com/mcp

Run this command in your terminal.

get_schemarun_query

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

Add to your MCP config in Windsurf settings.

get_schemarun_query

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

Add to your claude_desktop_config.json file.

get_schemarun_query

https://mcp.querybear.com/mcp

Paste this URL when adding an MCP server in ChatGPT.

get_schemarun_query

Common questions.

QueryBear currently supports PostgreSQL, MySQL, and SQLite. We're actively expanding to more databases — reach out if you need a specific one.

Your database credentials are encrypted at rest. Every query goes through a security layer with table allowlists, column-level privacy controls, and read-only enforcement. A full audit log tracks every query an agent runs.

No. You control exactly which tables and columns agents can see. QueryBear enforces read-only access by default, and your database credentials are never exposed to agents. We also have a desktop app coming soon — with it, database requests never leave your computer.

Any MCP-compatible client works out of the box — including Cursor, Claude Desktop, Claude Code, Windsurf, and ChatGPT. You just paste a single URL.

Absolutely. The Business plan supports team API key management so multiple agents and team members can share controlled access to the same databases. Admins get full RBAC controls to limit which tables, columns, and operations each team member or agent can access.