SDKs Overview
Official SDKs for Noukai, supporting Python and Node.js.
The Noukai SDKs let you execute AI workflows programmatically from your Python or Node.js applications. Both SDKs support streaming execution, tool handling, tracing, and error recovery with a unified API surface.
When to use Python vs Node
Python SDK (noukai-sdk):
- Data science pipelines and Jupyter notebooks
- FastAPI and Flask backends
- Batch processing and data jobs
- Scripts and REPL experiments
- Environments where you have both async and sync contexts
Node SDK (@noukai/sdk):
- Next.js and Remix routes
- Edge runtimes (Cloudflare Workers, Vercel Edge)
- Real-time browser-to-server bridges
- Async-first TypeScript backends
- Streaming responses to clients
Quick comparison
| Feature | Python | Node |
|---|---|---|
| Installation | pip install noukai-sdk | pnpm add @noukai/sdk |
| Sync API | ✓ (native + Noukai class) | ✗ (async only) |
| Async API | ✓ (AsyncNoukai class) | ✓ (default) |
| Streaming | flow.steps() / flow.events() | flow.steps() / flow.events() |
| Type hints | Full Pydantic | Full TypeScript |
| Min. version | Python 3.11+ | Node.js 18+ |
Getting started
Choose your language and follow the quickstart:
New to Noukai? Start with Quickstart, then explore Guides for deeper patterns.
- Python Installation — Install and configure the Python SDK
- Node Installation — Install and configure the Node SDK
- API Reference — Class signatures and method docs
Key features (both SDKs)
✨ Streaming execution — iterate step-by-step to see intermediate results 🔧 Tool handling — automatic retry logic for external function calls 🔍 Tracing — inspect cost, latency, and token usage per step ❌ Error recovery — typed exceptions with context-aware retry 🎛️ Versioning — execute specific flow versions (dev, production)
Integration
You can also use the HTTP API directly if you prefer low-level control, or the Noukai MCP server for AI-assisted development.