LibreChat, AnythingLLM, Lobe Chat
Multi-user, multi-model chat dashboards for self-hosted AI — alternatives and complements to Open WebUI.
After Open WebUI, this is the next tier of FOSS multi-user chat dashboards. Each has a slightly different center of gravity. For the broad picture see the overview and hardware tiers.
LibreChat
★ ★ LibreChat (github.com/danny-avila/LibreChat, MIT) — multi-provider chat dashboard.
- MIT license — more permissive than Open WebUI's AGPL.
- Multi-provider out of the box — OpenAI, Anthropic, Google, Mistral, Cohere, Ollama, vLLM, OpenRouter, Bedrock, Azure.
- Group chats — multiple AI participants in one thread.
- Plugins / agents with OpenAI Plugins and a custom tool framework.
- Code interpreter sandbox.
- Multi-user with auth, OAuth, LDAP, SAML.
- Conversation forking, search, presets.
- Active development since 2023; extremely feature-rich.
- ★ Best multi-provider story — if you switch between Claude, GPT, Gemini, and local Llama hourly, LibreChat is purpose-built for that.
AnythingLLM
★ AnythingLLM (github.com/Mintplex-Labs/anything-llm, MIT) — chat-with-your-documents focused.
- RAG-first design — workspaces are knowledge bases; chats are scoped.
- Many embedders / vector DBs supported — pgvector, Pinecone, Chroma, Weaviate, Qdrant, Milvus, LanceDB.
- Many LLM providers — Ollama, OpenAI, Anthropic, Mistral, OpenRouter, Groq, LM Studio, Together, etc.
- Agents with tool calling.
- Desktop app + Docker — both options.
- MIT license + paid hosted tier from Mintplex Labs for those who want managed.
- ★ The default if RAG is the headline feature — better RAG ergonomics than Open WebUI for serious document work.
Lobe Chat
★ Lobe Chat (github.com/lobehub/lobe-chat, Apache 2.0) — modern Next.js multi-provider chat.
- Slick UI; beautiful theming; Apple-iOS-flavour aesthetic.
- Plugin marketplace (Lobe Hub).
- PWA-installable.
- Multi-provider; voice; image gen; vision.
- Strong for individual / small-team aesthetic-conscious users.
Other notable
- Chatbot UI (github.com/mckaywrigley/chatbot-ui, MIT) — original FOSS ChatGPT clone; somewhat dormant in 2025–26 but still a clean simple base.
- AssistantUI (github.com/Yonom/assistant-ui) — React component library for embedding chat UIs in your own app rather than running a standalone one.
- NextChat (github.com/ChatGPTNextWeb/NextChat, MIT) — formerly ChatGPT-Next-Web; lightweight; popular in Asia.
Honest comparison
| Open WebUI | LibreChat | AnythingLLM | Lobe Chat | |
|---|---|---|---|---|
| License | AGPL-3.0 | MIT | MIT | Apache 2.0 |
| Multi-user / auth | ★ ★ | ★ ★ (LDAP/SAML/OAuth) | ★ | ★ |
| Multi-provider | ★ | ★ ★ | ★ | ★ |
| RAG strength | ★ | ★ | ★ ★ | ★ |
| MCP support | Yes | Partial | Partial | Limited |
| Group / multi-AI chat | Yes | ★ ★ | Yes | Limited |
| Tool / plugin system | ★ ★ | ★ ★ | ★ | ★ |
| Default-pick popularity | ★ ★ | ★ | ★ | ★ |
When to deploy multiple
It's normal for serious self-hosters to run more than one front-end. Open WebUI for daily chat, LibreChat for team multi-provider workflows, AnythingLLM for document-RAG. They all talk to the same Ollama / vLLM backends — point them at the same OpenAI-compat endpoint.
A common pattern: put LiteLLM in front of all your model providers, then point every front-end at LiteLLM. One source of truth for model routing, budgets, and logs.
Pick this if…
- Default chat UI: Open WebUI.
- Multi-provider central dashboard with enterprise auth: ★ ★ LibreChat.
- RAG over documents is the headline: ★ AnythingLLM.
- Aesthetic-first individual UI: Lobe Chat.
- Embedding chat into your own app, not running a standalone: assistant-ui or use the Vercel AI SDK.
- MIT-license preference (avoid AGPL): LibreChat or AnythingLLM.