Qaynaq

The fastest way to connect your data to AI

Connect any database, API, or service to AI assistants like Claude and Cursor. Open-source, runs on your machine, no coding required.

license Release Go Report Card

Website · Documentation · Playbooks

Why Qaynaq

Turn APIs, databases, and scripts into MCP tools without writing a single MCP server. Pick a connector, define your parameters, and get an instant endpoint your AI assistant can call.

  • Install in 30 seconds, build your first tool in 5 minutes
  • Your data stays on your machine - nothing leaves your network
  • 66+ connectors out of the box - databases, APIs, queues, and more
  • Free forever - open-source and self-hosted

Quick Start

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

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

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

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

Prefer Docker? docker pull ghcr.io/qaynaq/qaynaq. See the Installation docs for all options.

How It Works

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

Features

  • Instant MCP endpoint - flows auto-register as MCP tools, discoverable by any MCP client
  • External MCP server proxy - register remote (HTTP) or local (npx/command) MCP servers and serve their tools through the same /mcp endpoint, namespaced and aggregated, with OAuth and API key auth plus automatic token refresh
  • Visual tool builder - drag-and-drop DAG editor for tools and data flows
  • 66+ connectors - Kafka, HTTP, AMQP, MySQL CDC, PostgreSQL, and more
  • Built-in transformations - shape data with the Bloblang DSL and JSON Schema validation
  • Smart parameter validation - 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
  • Horizontal scaling - coordinator and worker architecture
  • Single binary - no Docker, JVM, or external dependencies required
  • Self-hosted and open-source - full control over your data, Apache 2.0 licensed

Playbooks

Contributing

We welcome contributions! Please check out CONTRIBUTING for guidelines.

License

Apache 2.0 - see LICENSE for details.