# Blokay > Blokay lets AI agents (Claude, ChatGPT, Cursor or any MCP client) query your production database safely. Every query runs with the permissions of the person who connected the agent: sensitive columns are masked or blocked, dangerous operations are rejected before they reach the database, AI queries are read-only, and everything is audited. Key facts: - MCP server URL: `https://app.blokay.com/mcp` (Streamable HTTP; OAuth 2.1 or a personal bearer token). - Supported databases: PostgreSQL, MySQL and MariaDB, also through an SSH tunnel. - Credentials are encrypted with AES-256-GCM and never reach the browser or the AI. - Plans: Free (1 account), Team (USD 8 per account per month), Enterprise (custom). See https://blokay.com/pricing - Contact: juan@blokay.com ## Introduction - [Give AI your production database, safely](https://blokay.com/docs/index.md): Blokay sits between your AI agent (Claude, ChatGPT, Cursor or any MCP client) and your database. The agent never gets a password: it signs in as a person on your team, and every query it sends goes through that person's permissions before it reaches your database. - [Quickstart](https://blokay.com/docs/quickstart.md): From zero to asking Claude about your production data, with your rules in place, in five steps. ## Connect your agent - [Connect Claude](https://blokay.com/docs/claude.md): Blokay is a remote MCP server. Claude connects to it with OAuth: you sign in to Blokay, pick a workspace, and Claude acts as **you**, with your permissions and nothing more. - [ChatGPT, Cursor and other agents](https://blokay.com/docs/agents.md): Any client that speaks MCP over HTTP can use Blokay. There are two ways to authenticate. - [MCP tools](https://blokay.com/docs/mcp-tools.md): These are the database tools an agent sees. Each one only appears if your role allows it. ## Databases - [Connect a database](https://blokay.com/docs/databases.md): Blokay connects to **PostgreSQL**, **MySQL** and **MariaDB**. ## Permissions - [Roles and access](https://blokay.com/docs/roles.md): Agents never have their own permissions: they inherit the ones of the person who connected them. So access starts with roles. - [Column and table restrictions](https://blokay.com/docs/restrictions.md): Restrictions decide what each person, and therefore each agent, can see inside a database. They are enforced by Blokay's executor before the query reaches your engine. - [Blocked operations](https://blokay.com/docs/operations.md): Some statements should never run against a database, no matter who asks. Blocked operations are rules **for everyone**, admins included, and they apply from every path: the explorer, Blokay's AI and MCP agents, in production and sandbox. ## Security - [Security](https://blokay.com/docs/security.md): Connecting a production database to an AI agent should not mean handing it a superuser password. This page explains each layer Blokay puts between the agent and your data. - [Audit and logs](https://blokay.com/docs/audit.md): When an AI agent touches production you need to answer three questions at any time: who ran it, what ran, and whether it was allowed. Blokay records it in two places. - [Two-factor authentication](https://blokay.com/docs/mfa.md): Production data deserves more than a password. Blokay supports two-factor authentication with any authenticator app (Google Authenticator, Authy, 1Password…). ## Optional - [Full documentation in one file](https://blokay.com/llms-full.txt) - [Pricing](https://blokay.com/pricing)