RAGFlow

What it is for

RAGFlow is a self-hosted RAG studio with an agent canvas: the official docs still (see the cover: RAG Customer Service, Begin, Intention Analysis, Tech Expert Agent on deepseek-chat, Tavily Search, Save / Run / Version history) is the graph. Use it when the job is chunk documents and route an agent over them, not when you want a generic chat shell.

The software is Apache-2.0. Self-host is $0. We found no public RAGFlow Cloud dollar on ragflow.io or cloud.ragflow.io.

What makes it unique: the cover is InfiniFlow’s own agent chrome — Begin → Intention Analysis → expert agents + Tavily — not a logo tile.

Key features

• Agent canvas — Begin, Intention Analysis, Tech / Product expert agents, Retrieval, Tavily Search (cover).
• Save / Run / Version history / Management — toolbar on the official GIF.
• Template-based chunking — official README stills show explainable chunks for PDFs, slides, and tables.
• Knowledge-base file list — enable/disable parsers per file (official carousel still).
• Self-host + Cloud — Docker on the README; Cloud CTA at cloud.ragflow.io (no printed seat).

Who it's for

Best fit: teams who will run Docker and already have a chat-model key (the cover uses DeepSeek).

Poor fit: people who want a ChatGPT-style shell only (Open WebUI) or a hosted Dify card (Dify).

Pricing

Verified 2026-10-21 (Asia/Shanghai) from github.com/infiniflow/ragflow (Apache-2.0) and ragflow.io.

• Self-host / Docker — $0. Apache-2.0. You bring CPU/GPU, disk, and model keys.
• Hosted Cloud SKU — not on the pages we fetched. Do not invent one.

Limitations / costs

• $0 is the software, not Tavily or DeepSeek. The cover’s search and chat nodes still bill their vendors.
• First-run chunking is heavy. PDFs and tables need RAM and time.
• No printed Cloud card. If one appears, re-check ragflow.io.
• Not a general assistant. The product is RAG + agents, not a multi-model bot shelf.

Alternatives

• Dify — visual LLM studio with a public Cloud Professional card.
• Langflow — MIT agent canvas; you bring the key.
• AnythingLLM — local workspace over files; MIT.

Official tutorial

Official RAGFlow channel video (linked from ragflow.io): RAGFlow: Build a Superior Context Layer for Your AI Agents (oEmbed author: RAGFlow).

Written README Get Started plus Documentation.

Website

https://ragflow.io

Repository

https://github.com/infiniflow/ragflow

Source

Catalog row curated for AI Tools Nav. Cover is InfiniFlow’s official agentic-dark GIF (RAG Customer Service canvas), not the wordmark. License Apache-2.0. Tutorial: Build a Superior Context Layer on the official RAGFlow channel (oEmbed).

Last verified

2026-10-21 (Asia/Shanghai) editorial check. Apache-2.0 self-host is $0. No Cloud dollar on the fetched pages.