Skip to main content
Ragen AI is a self-hosted, open-source RAG (Retrieval-Augmented Generation) platform that lets your team and customers ask questions of your company’s documents and get answers grounded in your actual data — with every answer traceable back to the source. You run it entirely on your own infrastructure: your documents, your database, your vector index, your encryption keys.

Quick Start

Scaffold a working Ragen instance in under 10 minutes with a single command.

Self-Hosting Guide

Set up Ragen on your own infrastructure with Docker Compose or Kubernetes.

API Reference

Integrate Ragen into your apps with the OpenAI-compatible REST API.

TypeScript SDK

Use the official @webamigos/ragen-sdk-ts client for typed, streaming-ready access.

​
What is Ragen AI?

Ragen connects your existing documents — PDF, DOCX, PPTX, XLSX, CSV, Markdown, images, URLs, and more — to modern language models using Retrieval-Augmented Generation. When a user asks a question, Ragen retrieves the most relevant document chunks and generates an answer that cites those chunks by source, so your team always knows where the information came from.

​
Key capabilities

AI Chat Assistant

Internal knowledge platform for your team, with document permissions enforced at retrieval — not just in the UI.

Embeddable Chatbot

Customer-facing widget you can embed on any website, backed by your own knowledge base.

OpenAI-Compatible API

Drop-in replacement for OpenAI’s Chat Completions, Files, Assistants, and Threads APIs.

No Model Lock-In

Route to OpenAI, Anthropic, Azure, AWS Bedrock, Google Vertex, or models running on your own hardware.

Multilingual Retrieval

Hybrid dense + sparse (BM25) search with cross-encoder reranking. 92% multilingual benchmark accuracy.

MCP Integrations

Connect Google Drive, Google Calendar, Gmail, HubSpot, ClickUp, and Slack as live data sources.

​
How it works

1

Upload your documents

Add PDFs, Word docs, spreadsheets, web pages, or any supported file to a project (knowledge base). Ragen parses, chunks, and embeds them automatically.
2

Ask a question

Users ask questions in the chat interface, via the API, or through an embedded chatbot. Ragen retrieves the most relevant document chunks using hybrid vector + keyword search.
3

Get a grounded answer

The language model generates an answer using only the retrieved content, and Ragen shows which documents and passages it cited — so you can verify every response.

​
Deployment options

A demo instance is available at demo.ragen.ai seeded with sample data. It shows what a Ragen installation looks like — your own data belongs on your own instance.

​
Next steps