What is RAGBot?

RAGBot lets you chat with your own documents. Upload the files you already rely on, and a set of AI assistants can read them and answer your questions — in plain language, with links back to the exact pages they pulled from.

Instead of searching through folders or scrolling a 90-page PDF, you just ask. Behind the scenes RAGBot finds the relevant passages across your files and writes an answer grounded in them, so you can trust what you read and check the source yourself.

A quick word on "RAG"

RAG stands for retrieval-augmented generation — a fancy way of saying the assistant looks things up in your documents before it answers, rather than guessing from memory. You never have to think about it. You just upload, then ask.

The four building blocks

Everything in RAGBot is made of four simple pieces. Once these click, the rest of the product follows naturally.

Workspaces

A sealed space that holds all your datasets, agents, chats, and people. Nothing is shared between workspaces.

Datasets

Collections of uploaded documents — PDFs, Word files, text. RAGBot indexes them so agents can search inside.

Agents

Assistants you configure with a model and a persona. The workspace ships a ready-to-use default agent.

Chat

Pick an agent, choose which sources to search, ask a question, and read answers with cited sources. History is saved.

How it fits together

The pieces stack in one direction. A workspace contains everything. Inside it, you gather documents into datasets. To get answers, you open a chat: pick an agent (its persona and model) and choose which datasets it should search. The agent decides how to answer; the chat decides what it reads.

Put simply: workspace › datasets, then chat = agent + sources. If you can remember that, you can find your way around the whole product.

New here?

The fastest way to understand RAGBot is to build a tiny example yourself. The Quick Start walks you from an empty workspace to your first answered question in five short steps.