FastGPT

What it is for

FastGPT is a self-hosted RAG/workflow studio: the official README still (see the cover: cloud.fastgpt.io, 智能助手, 应用配置 / 发布渠道 / 对话日志, 运行 / 保存, node graph with reply, Feishu send, text, and code nodes) is the canvas. Use it when the job is draw a knowledge workflow and publish a channel, not when you want a single chat box.

The software is Apache-2.0 with additional conditions (no unauthorized multi-tenant SaaS clone; keep the logo). Self-host is $0. We found no public FastGPT Cloud dollar on fastgpt.io.

What makes it unique: the cover is FastGPT’s own workflow chrome — Run/Save plus Feishu and code nodes — not a wordmark.

Key features

• Workflow canvas — reply, Feishu, text, and code nodes with Run / Save (cover).
• Debug / publish / chat logs — tabs on the official still.
• Knowledge-base chat — official intro1 still shows a cited FastGPT assistant.
• Search test — official intro3 still: semantic retrieval with scores.
• Self-host — Apache-2.0 core. Cloud hostname on the cover is cloud.fastgpt.io.

Who it's for

Best fit: Chinese/English teams who will self-host a knowledge assistant and already have a model key.

Poor fit: people who want an English-only hosted studio with a printed card (Dify) or a simple local chat (Open WebUI).

Pricing

Verified 2026-10-21 (Asia/Shanghai) from github.com/labring/FastGPT (LICENSE: Apache-2.0 plus the two extra conditions) and fastgpt.io.

• Self-host — $0. Apache-2.0 with additional conditions. You bring the model key and a database.
• Hosted Cloud SKU — not on the pages we fetched (/pricing 404s). Do not invent one.

Limitations / costs

• $0 is the software, not the LLM. Tokens still bill the vendor.
• Additional license conditions. Multi-tenant SaaS clones and logo removal need a commercial license (dennis@sealos.io).
• UI in the official stills is Chinese-first. English copy on this page is ours.
• No printed Cloud $. If a card appears, re-check fastgpt.io.

Alternatives

• Dify — visual studio with a public Cloud Professional card.
• RAGFlow — Apache-2.0 RAG + agent canvas.
• Langflow — MIT visual agents.

Official tutorial

Quick Start walks through local, Docker, and Sealos install. Official team courses live on FastGPT 视频中心 (新手教学 / 部署教程 / 功能介绍).

Website

https://fastgpt.io

Repository

https://github.com/labring/FastGPT

Source

Catalog row curated for AI Tools Nav. Cover is FastGPT’s official README workflow still (cloud.fastgpt.io canvas), not a logo. License Apache-2.0 with additional conditions. Written tutorial: Quick Start. Official courses: video.fastgpt.cn/videos.

Last verified

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