Qaynaq

The fastest way to connect your data to AI

Turn APIs, databases, and scripts into MCP tools - without writing MCP servers. Qaynaq is an open-source, self-hosted platform that lets you visually build tools for AI assistants like Claude, Cursor, and AI agents. Connect 66+ sources, define parameters, and get an instant MCP endpoint. Single binary, no Docker or JVM required.

Documentation | Playbooks

Quick Start

Install

curl -Lsf https://qaynaq.io/sh/install | bash

See the Installation docs for more options.

Run

# Start coordinator
qaynaq -role coordinator -grpc-port 50000

# Start worker (in a separate terminal)
qaynaq -role worker -grpc-port 50001

Open http://localhost:8080 and start building MCP tools.

Docker

docker pull ghcr.io/qaynaq/qaynaq

See the Installation docs for Docker run commands and the Configuration docs for database setup.

How It Works

1. Connect - Pick from 66+ built-in connectors for APIs, databases, queues, and services.
2. Define Tool - Set the tool name, description, and input parameters. Qaynaq generates the MCP tool automatically.
3. Use With AI - Tools are instantly callable by Claude, Cursor, AI agents, and internal copilots via the /mcp endpoint.

Features

• Instant MCP Endpoint - Flows are auto-registered as MCP tools, discoverable by any MCP client
• Visual Tool Builder - Drag-and-drop DAG editor to build and manage tools and data flows
• 66+ Connectors - Kafka, HTTP, AMQP, MySQL CDC, PostgreSQL, and more
• Built-in Transformations - Transform data with Bloblang DSL and JSON Schema validation
• Parameter Validation - Define types, descriptions, and required flags visible to AI assistants
• Secure Credentials - Encrypted secrets management for API keys and tokens
• Automation Flows - Move and route data between systems, or orchestrate AI workflows that call your MCP tools
• Horizontal Scaling - Coordinator & worker architecture for easy scaling
• Single Binary - No Docker, JVM, or external dependencies required
• Self-hosted & Open Source - Full control over your data, Apache 2.0 licensed

Playbooks

• Build a Weather Tool for AI Assistants - Build tools for AI assistants without writing MCP servers.
• Kafka to PostgreSQL - Flow events from Kafka through Avro schema decoding into PostgreSQL.
• HTTP Webhooks - Accept webhook data over HTTP and store it in a database.

Contributing

We welcome contributions! Please check out CONTRIBUTING (coming soon) for guidelines.

License

Apache 2.0 - see LICENSE for details.