What is Haystack?

Haystack is an open-source AI orchestration framework built by deepset that empowers Python developers to build real-world, compound, agentic LLM applications.

Whether you’re building autonomous agents, multi-modal apps, or scalable RAG systems, Haystack gives you the tools to go from idea to production with ease. Its powerful, modular architecture helps you try out state-of-the-art methods, customize deeply when needed, and stay in control of your stack.

Backed by an inspiring community of developers, Haystack has evolved into a complete and intuitive framework for LLM-based development.

pip install -U haystack-ai

📘 Want to learn more how to use Haystack? Check out the docs

Who is Haystack For?

Haystack is for builders of all kinds—whether you’re prototyping your first Gen AI app or deploying sophisticated systems at scale. If you know a bit of Python, you’re ready to start.

You can build directly in Python, visually with deepset Studio, or scale up with enterprise-grade infrastructure on the deepset AI Platform. Whatever your level or need, Haystack gives you the power to build, experiment, and launch faster.

Why Developers Love Haystack

Built for production from day one, Haystack is the leading open-source framework for building reliable, flexible, and composable LLM applications from first prototype to full-scale deployment.

Here’s why developers choose Haystack:

• Build flexible pipelines using Haystack’s built-in components like retrievers, generators, rankers, agent, or create your own.
• Customize with full control by adding loops and branches to your pipeline, and tailor data flow to fit your use case.
• Integrate smoothly with model providers, vector databases, and AI tools in a modular, swappable architecture
• Deploy effortlessly and serve your pipelines as REST APIs or as an MCP Server using with Hayhooks.
• Monitor and optimize using built-in tracing, logging, and evaluation tools for full observability.
• Rely on robust engineering with clean architecture, careful dependency management, and a breaking change policy that ensures no bad surprises.

Whether it’s agent workflows, retrieval over millions of docs, text-to-SQL or PDF parsing, Haystack scales with your idea without scaling your complexity.

The deepset Ecosystem

At deepset, we’re building the tools to make LLMs truly usable in production—open source and beyond.

• Haystack → Our core framework for LLM orchestration, launched in 2020 and now at v2.
• deepset Studio → A free visual interface for building pipelines, experimenting with data, and deploying apps faster.
• deepset AI Platform → An enterprise-ready platform for teams running Gen AI apps in production, with security, governance, and scalability built in.

Recognized as a 2024 Gartner Cool Vendor in AI Engineering, we’re proud to serve the community and enterprises alike.

Haystack Community

Behind Haystack is a global community of makers, researchers, and tinkerers who believe in the power of open-source collaboration. From first-time contributors to core developers, our community brings a rich mix of backgrounds, interests, and skills.

We’d love to have you involved:

• 🤝 Connect with others and ask questions on Discord
• 🧑‍💻 Contribute to the project on GitHub
• 🎙️ Join us at live events, webinars, and office hours
• 📺 Catch talks and tutorials on YouTube
• 🐘 Follow along on X (Twitter) or Mastodon
• 💌 Get our latest news in your inbox via the Haystack newsletter

Enter the Haystack Universe

There are so many ways to start building:

• Start building with tutorials
• Follow use case examples in cookbook
• Learn with the DeepLearning.AI course
• Check out integrations
• Explore cutting-edge features in Haystack Experimental
• Read our blog and docs