Coremcp, developed by Corebasehq, is a free Model Context Protocol (MCP) server designed to connect AI agents to SQL databases. Built with Go, it functions as a secure, tunnel-native bridge, enabling AI agents to access legacy and on-premise databases behind firewalls. It supports MSSQL and PostgreSQL, with other database adapters under development. Coremcp ensures database credentials remain on the user's server, providing a secure method for AI agent data interaction. It is a key component for teams building AI features that require data access.

What does Coremcp connect to?

Coremcp connects to SQL databases, specifically supporting MSSQL (SQL Server 2000+) and PostgreSQL adapters. Firebird support is in progress, and MySQL is on the roadmap. It acts as a secure, tunnel-native database bridge, allowing AI agents to interact with legacy and on-premise databases that are located behind a firewall. This design ensures that internal data sources can be safely exposed to external AI systems without compromising security.

Core features and tools

Coremcp exposes specific callable tools including custom tools and schema discovery. Its relay protocol supports wire commands such as run_sql, get_schema, list_sources, health_check, and config_sync. When configured with "readonly: false", it also enables the execute_procedure action. Additionally, Coremcp provides a database_schema prompt, which supplies schema context directly to models, aiding AI agents in understanding database structures for more accurate queries.

Who is it for?

Coremcp targets teams developing AI features for their products, especially those needing to connect AI agents to their company's data, including legacy and on-premise systems behind a firewall. It serves developers who need to bridge internal SQL databases to AI platforms securely. It is ideal for organizations prioritizing data security while integrating AI capabilities. Skip this if your AI agents do not need to connect to on-premise or legacy SQL databases.

How to get started

Access is granted through a standard OAuth authorization flow for remote operation. For remote operation using the connect mode via WebSocket, it requires an authentication --token.

Use cases

• Connecting AI agents to on-premise SQL databases: It acts as a secure bridge, allowing AI agents to query and analyze data from SQL Server and PostgreSQL databases located behind a firewall.
• Enabling secure data analysis by AI clients: It supports interaction between AI clients (like Claude Desktop) and local databases, incorporating safety features such as read-only by default, a T-SQL aware lexer to prevent malicious queries, forced row caps, and PII masking.
• Integrating with the CoreBase AI agent platform: Coremcp connects to CoreBase Cloud via an outbound WebSocket, allowing CoreBase agents to access internal systems and execute multi-step automations across various data sources.

Coremcp provides a secure and structured way to link AI agents with internal SQL databases, supporting both local and remote operations. Its focus on security, with features like read-only defaults and PII masking, makes it a reliable choice for sensitive data interactions. The token-based authentication for remote access ensures controlled connections.