CentralMind Gateway provides a universal server for exposing structured databases to AI agents and LLM-powered applications. It automatically generates secure, LLM-optimized APIs via protocols like MCP or OpenAPI, enabling efficient data access, PII filtering, and audit trails for AI workloads.

How It Works

The gateway connects to databases, analyzing schema and data samples. It leverages LLMs to generate an API configuration, incorporating PII detection for security. This configuration is then deployed as a server, accessible via REST or MCP protocols, with features like caching and row-level security for optimized AI integration.

Quick Start & Requirements

• Install/Run: docker run --platform linux/amd64 -p 9090:9090 ghcr.io/centralmind/gateway:v0.2.14 start --connection-string "postgres://..."
• Prerequisites: Docker, PostgreSQL (or other supported databases).
• Setup: Quick start involves a single Docker command.
• Docs: Getting Started, API Generation

Highlighted Details

• Supports PostgreSQL, MySQL, ClickHouse, Snowflake, BigQuery, Oracle, MSSQL, Elasticsearch, and SQLite.
• Offers REST and MCP (SSE) API protocols with auto-generated Swagger/OpenAPI 3.1.0 documentation.
• Includes PII redaction via regex or Microsoft Presidio plugins, and row-level security with Lua scripts.
• Integrates with OpenAI, Anthropic, Bedrock, Gemini, and VertexAI, supporting self-hosted LLMs.
• Features OpenTelemetry integration for monitoring and caching strategies for performance.

Maintenance & Community

• Active development indicated by recent commits and releases.
• Roadmap available, focusing on extended database support, write operations, and advanced traffic management.

Licensing & Compatibility

• License: Apache 2.0.
• Compatibility: Suitable for commercial use and integration with closed-source applications.

Limitations & Caveats

The project is currently focused on read-only operations, with write support planned for future releases. The LLM-based API generation relies on external AI providers, and the quality of generated APIs may vary.