Skip to main content
For the complete documentation index for agents and LLMs, see llms.txt.

What's Haystack Enterprise Platform?

Haystack Enterprise Platform is a platform for building AI agents and retrieval-augmented generation (RAG) apps and running them in production. It covers the full lifecycle: connecting your data, building and testing pipelines, deploying them behind an API, and monitoring them once they're live.

Haystack Enterprise Platform is also the platform for sovereign AI: you decide where your apps run, who operates the infrastructure, and which models they use. The platform is built on Haystack, deepset's open source framework, so pipelines you build here use the same components and structure as Haystack.

Connect and Index Your Data​

Upload files or connect enterprise systems such as SharePoint and Google Drive. Built-in converters and preprocessors handle PDFs, Office documents, HTML, images, and audio. Configure an index once, and the platform cleans, chunks, and stores your data in a document store, ready for your pipelines to retrieve.

Use the managed OpenSearch document store, or bring your own, such as Elasticsearch, Pinecone, Qdrant, or Weaviate. For details, see Integrate with Enterprise Systems and Document Stores.

Build Agents and RAG Pipelines​

Build AI agents that reason, call tools, and complete multi-step tasks, or RAG pipelines that ground answers in your documents. You design pipelines in Builder, a visual editor where you drag and drop components or edit YAML directly. Start from a template, get help from the AI Assistant when something doesn't work, and add your own Python code when you need custom logic.

To learn more, see Building AI Agents and Create a Pipeline in Studio.

Use Any Model​

Haystack Enterprise Platform is model-agnostic. Use models from OpenAI, Anthropic, Amazon Bedrock, Google Gemini, Mistral, NVIDIA, and other providers, or models you host yourself. You can combine several models in one app and switch models without changing your pipeline. Save custom model definitions to reuse your preferred settings across pipelines.

For the full list, see Supported Connections and Integrations.

Test and Improve with Real Users​

Try your pipeline in the Playground and refine prompts in Prompt Explorer, comparing responses side by side. When it's ready for feedback, share a branded prototype link that testers can open without an account, and collect their ratings and comments.

For details, see Share a Pipeline Prototype and Collect User Feedback.

Deploy and Integrate​

Deploy a pipeline in one click. Deployed pipelines scale automatically with traffic, and pipeline versions let you test changes before you make them the default. Call your pipelines from your own apps through the REST API or the Python SDK, stream responses as they're generated, or expose pipelines as tools for MCP clients such as Cursor and Claude Code.

For details, see Deploy a Pipeline and Use Your Pipelines as MCP Tools.

Monitor and Debug​

See what your pipelines do in production. Built-in traces show each run step by step, with inputs, outputs, timings, and errors. Pipeline logs, search history, and performance dashboards help you spot problems and track usage. You can also send traces to Langfuse, Weights & Biases Weave, or your own OpenTelemetry or Datadog stack.

For details, see Trace with Built-In Traces and Monitor Pipeline Performance.

Run Your AI Where You Need It​

Run Haystack Enterprise Platform as a managed cloud service, or self-host it in your private cloud, in your on-premises data center, or in an air-gapped environment with no internet connection. With a self-hosted deployment, your files, document stores, traces, and models stay in infrastructure you control, so you can meet data residency and sovereignty requirements. For details, see Deployment Options.

Keep Control with Security and Governance​

Organize teams into isolated workspaces and control access with preset or custom roles. Store API keys as secrets instead of in your pipelines, and set limits on what each agent can do, such as the tools it can call and how many steps it can take. deepset is SOC 2 Type II and ISO 27001 certified.

For details, see Enterprise Security and Governance and Agent Governance and Control.

Where to Start​