the catalog · ✨ AI assistants

AnythingLLM logo

AnythingLLM

★ 66k

Private AI workspace for chatting with documents, running agents, and swapping local or cloud LLMs without sending your files out.

AnythingLLM does the job of ChatGPT Team and Custom GPTs, self-hosted. It is MIT-licensed, so you can do essentially what you like with it, and it installs in about one command. ChatGPT Team bills $25+ for it. 66k stars on GitHub.

Open-source replacement for ChatGPT Team, Custom GPTs.

Licence
MIT
Self-hosting
Easy
The paid one costs
$25+
GitHub stars
66k
view the code →official site →

All 18 paid apps AnythingLLM replaces, and the catch with each →

Is it actually a replacement?

We have written a verdict on what it replaces — what you lose, and why people still pay:

How AnythingLLM runs

Where it runs

AnythingLLM runs on your machine or their cloud.

Installing it

Install the one-file Windows, macOS, or Linux desktop app, or run the self-hosted Docker image with a persistent storage volume.

Where your data lives

Desktop data lives in a local AnythingLLM storage folder containing anythingllm.db, parsed documents, vector indexes such as LanceDB, caches, and optional GGUF models; Docker uses the mounted storage volume, while external providers receive retrieved context.

What setup takes

The app installs cleanly, but useful private operation still requires choosing and downloading a model that fits the machine; the server route also requires Docker and correct volume mounting.

The catch compared with Afforai

It can search and cite local files, but setup, citation checking, side-by-side source comparison, and a managed research workspace require more work than Afforai.

AnythingLLM questions

Is AnythingLLM open source?

Yes — it is released under MIT.

Can AnythingLLM replace ChatGPT Team?

Our verdict on ChatGPT is YES. Dify (156k stars), RAGFlow (91k stars), Lobe Chat (82k stars) are mature open-source replacements for ChatGPT — self-host with docker and stop paying $20/mo.

Can I self-host AnythingLLM?

Yes — we rate it easy to self-host, and it runs on your machine or their cloud. Install the one-file Windows, macOS, or Linux desktop app, or run the self-hosted Docker image with a persistent storage volume.

More local model runner tools