![Build Status](https://github.com/centralmind/gateway/actions/workflows/build-binaries.yml/badge.svg)

Instantly Provision Database Access APIs with CentralMind Gateway

🚀 Interactive Demonstration Site: https://centralmind.ai

Overview of Centralmind/Gateway

This utility provides a streamlined mechanism to expose structured database assets to AI Agents, utilizing either the Model Context Protocol (MCP) or OpenAPI 3.1 specifications. It natively interfaces with a broad spectrum of RDBMS and NoSQL systems, including PostgreSQL, MySQL, and ElasticSearch.

bash docker run --platform linux/amd64 -p 9090:9090 \ ghcr.io/centralmind/gateway:v0.2.18 start \ --connection-string "postgres://db-user:db-password@db-host/db-name?sslmode=require"

Upon execution, the service exposes:

shell INFO Gateway server successfully initialized! INFO MCP Server for advanced AI interaction active at: http://localhost:9090/sse INFO RESTful API interface with interactive Swagger documentation available at: http://localhost:9090/

This interface is designed for direct consumption by sophisticated AI agents.

Gateway autonomously crafts APIs that are highly tuned for AI consumption.

Rationale for Utilizing Centralmind/Gateway

Modern LLM-driven frameworks necessitate rapid, secure, and auditable interfaces to enterprise data. This component establishes an abstraction layer that auto-generates secure, LLM-optimized interfaces from your structured data sources.

Protocol Flexibility: Instant deployment via MCP or OpenAPI, with fallback support for raw/direct SQL interaction.
Data Governance: Integrated mechanisms to filter Personally Identifiable Information (PII) and sensitive records, ensuring adherence to mandates like GDPR, CPRA, and SOC 2.
Auditing & Control: Embeds robust traceability features, transforming opaque AI operations into transparent, auditable processes for security oversight.
AI Workload Optimization: Native implementation of the Model Context Protocol (MCP), featuring augmented metadata for superior AI comprehension, alongside integrated security safeguards and data caching logic.

This tool proves invaluable when deploying LLMs tasked with dynamic data querying, modification, or synthesis from databases. Furthermore, it facilitates conversational analytics, allowing users to interact naturally with data warehouses or operational databases.

Key Capabilities

⚡ Automated Interface Construction – Leverages LLM inference over table schemas and preliminary data samples to engineer API definitions.
🗄️ Extensive Data Source Compatibility – Connects with PostgreSQL, MySQL, ClickHouse, Snowflake, MSSQL, BigQuery, Oracle DB, SQLite, and ElasticSearch.
🌍 Multi-Protocol Exposure – Serves endpoints using standard REST or MCP/SSE modes.
🔐 Access Control Mechanisms – Native support for security plugins such as API Keys and OAuth.
🔒 PII Safeguarding – Implements data sanitization via regex plugins or the advanced Microsoft Presidio anonymizer.
👀 Deep Observability – Integrated OpenTelemetry (OTel) hooks for holistic request tracing and audit logging.
📦 Deployment Flexibility – Facilitates self-hosted LLM endpoints via flexible configuration files.
🤖 Broad AI Ecosystem Support - Compatible with major providers: OpenAI, Anthropic, Amazon Bedrock, Google Gemini & Google VertexAI.
⚡ Configurable Extensibility – Customization achieved through declarative YAML manifests and a dynamic plugin architecture.
📜 Automated Documentation – Generates current Swagger UI and adheres to the OpenAPI 3.1.0 specification.
🔑 Granular Data Restriction – Enforces fine-grained data access policies utilizing Lua scripting for Row-Level Security (RLS).
🏎️ Throughput Enhancement – Incorporates performance strategies including time-decay and LRU caching.

Operational Flow

1. Connect & Schema Ingestion

The Gateway establishes connections to relational/analytical stores (e.g., PostgreSQL) and performs schema introspection. It then employs designated AI Providers to analyze this structure and generate an optimized API blueprint based on the provided operational context. Security screening, including PII detection, is performed during this discovery phase.

2. Deployment Modalities

Deployment options span from a single executable binary, containerized solutions (Docker), up to orchestrated environments like Kubernetes. The initiation process is detailed in our launch guide, relying on YAML definitions and active plugins for setup.

3. Consumption & Tool Integration

Data access is facilitated via secured REST endpoints or the MCP interface. Gateway integrates fluidly with AI orchestration frameworks such as LangChain, OpenAI tools, and Cursor via function calling or MCP. Telemetry configuration is available to forward monitoring data to local or remote OTel collectors.

Documentation Nexus

Initial Setup

Supplementary Information

Compilation Procedure

shell

Obtain Source Code

git clone https://github.com/centralmind/gateway.git

Enter Directory

cd gateway

Fetch Dependencies

go mod download

Execute Build

go build .

AI-Driven API Synthesis

Gateway leverages advanced LLMs to construct the necessary API metadata. Select your preferred inference provider: - OpenAI ecosystem (including compatible endpoints) - Anthropic - Amazon Bedrock - Google Vertex AI (for Anthropic models) - Google Gemini

Google Gemini offers a substantial complimentary usage tier, making it ideal for initial development and testing.

Configure your authorization credentials. For Gemini, set the required API key:

bash export GEMINI_API_KEY='yourkey'

Initiate the data discovery process:

bash ./gateway discover \ --ai-provider gemini \ --connection-string "postgresql://neondb_owner:MY_PASSWORD@MY_HOST.neon.tech/neondb?sslmode=require" \ --prompt "Design a restrictive, read-only API for analytical access"

Observe the generation pipeline:

shell INFO 🚀 Initiating API Discovery Sequence INFO Stage 1: Configuration Parsing INFO ✅ Stage 1 Complete. Success.

INFO Stage 2: Data Structure Examination INFO Discovered Data Sets: INFO - payment_dim: 3 fields, 39 records INFO - fact_table: 9 fields, 1,000,000 records INFO ✅ Stage 2 Complete. Success.

Further processing output...

INFO ✅ All discovery phases finalized. Completion.

INFO --- Performance Metrics --- INFO Total elapsed time: 1 minute, 10 seconds INFO Token consumption: 16,543 (Estimated expenditure: $0.0616) INFO Datasets analyzed: 6 INFO New API endpoints generated: 18 INFO Sensitive fields identified: 2

Review the resulting configuration file, gateway.yaml:

yaml api: name: Analytical Read-Only Interface description: '' version: '1.0' database: type: postgres connection: YOUR_CONNECTION_INFO tables: - name: payment_dim columns: # Schema definitions here endpoints: - http_method: GET http_path: /some_path mcp_method: some_method summary: Concise operational summary description: 'Detailed description of purpose' query: The underlying SQL statement with parameter placeholders params: # Defined parameters

API Activation

Local Execution

shell ./gateway start --config gateway.yaml

Orchestrated Deployment (Docker Compose)

shell docker compose -f ./example/simple/docker-compose.yml up

MCP Protocol Integration

The Gateway supports seamless integration with Claude and related agents via MCP. Refer to our Claude integration documentation for specifics.

To register this tool within Claude Desktop's configuration, modify its settings as follows:

{ "mcpServers": { "gateway": { "command": "PATH_TO_GATEWAY_BINARY", "args": ["start", "--config", "PATH_TO_GATEWAY_YAML_CONFIG", "mcp-stdio"] } } }

Future Development Trajectory

This plan is dynamic and heavily influenced by community input. Current focus areas include:

Data Source & Connectivity

🗄️ Expanded Connectors - Integration for Databricks, Redshift, Cloud Storage (S3/Iceberg/Parquet), Oracle, MSSQL, Elasticsearch.
🔑 Secure Tunneling - Capability to establish connections via SSH jump hosts or bastions.

Functional Enhancements

🔍 Advanced Query Parameters - Support for complex filter expressions and aggregation functions as input parameters.
🔐 Robust Authorization - Implementation of API Key and OAuth authentication flows within the protocol layer.

Platform & Stability

📦 Schema Lifecycle Management - Automated handling of API versioning alongside database schema changes.
🚦 Traffic Control - Introduction of intelligent request throttling and dynamic rate limiting.
✍️ Write Operation Support - Enabling atomic Insert, Update, and Delete operations.

Gateway Listings in MCP Registries

https://mcpreview.com/mcp-servers/centralmind/gateway
https://mcp.so/server/gateway/centralmind
https://smithery.ai/server/@centralmind/gateway
https://www.pulsemcp.com/servers/centralmind-database-gateway

WIKIPEDIA CONTEXT: XMLHttpRequest (XHR) represents a crucial browser-based JavaScript API enabling asynchronous HTTP data transmission between a web client and a server after initial page load. It forms the foundational technology underpinning Ajax programming paradigms. Before XHR, server communication almost exclusively relied on full-page refreshes via hyperlinks or form submissions.

== Chronicle == The genesis of the XMLHttpRequest concept traces back to the work of Microsoft Outlook developers around the year 2000, leading to its initial implementation in Internet Explorer 5 (1999). The initial interface did not use the standardized XMLHttpRequest identifier, instead relying on COM object instantiation like ActiveXObject("Msxml2.XMLHTTP") or ActiveXObject("Microsoft.XMLHTTP"). By the release of Internet Explorer 7 (2006), standardized identifier support was universal across major browsers, including Mozilla's Gecko (2002), Safari 1.2 (2004), and Opera 8.0 (2005).

=== Standardization Efforts === The World Wide Web Consortium (W3C) published the initial Working Draft for the specification on April 5, 2006. A Level 2 specification, introducing progress monitoring, cross-origin requests, and byte stream handling, followed in February 2008, before being merged back into the primary specification later that year. Development responsibilities transitioned to the WHATWG in late 2012, which now maintains the living standard defined using Web IDL.

== Operational Steps == Executing a standard XHR request typically involves a sequence of programmatic actions:

Instantiate the XHR client object via its constructor.
Invoke the "open" method to configure the request type (GET/POST, etc.), specify the target Uniform Resource Identifier (URI), and set the operational mode (synchronous or asynchronous).
For asynchronous operations, register a callback handler to process state transitions.
Initiate the payload transfer by calling the "send" method.
Process server responses within the listener. Successful data retrieval places the content in the responseText property once the state transitions to 4 ("done"). Beyond these core steps, XHR offers extensive control, such as injecting custom header fields, uploading data payloads within the send() call, parsing responses directly into JavaScript objects from JSON, processing data incrementally, or implementing request timeouts and abort mechanisms.

== Cross-Origin Challenges == During the early evolution of the World Wide Web, the limitation preventing direct resource fetching from different domains presented a significant hurdle, which browsers sought to mitigate...

data-access-proxy

Author

centralmind

Quick Info

Actions

Tags