A self-hosted AI chat interface gives your team a familiar chat window on your own server, but it's only the window. None of these ships its own model: you connect a local model or an API key, and your privacy depends on where that model runs. Pick by who'll use it, then check the licence.
Want a chat window for your staff that looks like the big AI assistants but runs on a server you control? That's what these projects offer. Before choosing one, it helps to know exactly what you're getting, and what you're not.
Below are eight open front ends grouped by who they're built for: whole teams with logins, one person or a small group, and people who mainly want to chat with their own documents. They're listed in no particular order, and none is crowned.
The window and the model are two separate things
Every tool on this page is the interface: the chat box, the conversation history, the user accounts and settings. None of them includes an AI model. You plug one in, either a model running on your own hardware through a runner such as Ollama, or a hosted model reached with an API key.
That's the part people miss. If you connect a hosted model's API, your prompts still travel to that provider, however private the chat window looks. If you run the model locally, nothing leaves your network, but you'll need hardware that can run it. Choose the model side first, then the window.
Which front end for which kind of user
Team workspaces with logins
These are built for an organisation: user accounts, shared settings and several model connections. LibreChat brings several AI providers, agents and tools into one multi-user app, and its site says the project is joining ClickHouse. The other two in this group are a self-hosted workspace designed to run entirely offline and an agent workspace with Docker Compose deployment, each with licence conditions covered below.
Light clients for one person or a small group
These deploy quickly and suit personal use. NextChat is a light client usually deployed through Vercel or Docker, with an enterprise private-deployment offer on the side. Big-AGI lets you compare and combine answers from several models, and its hosted product is built on the open repo. Hugging Face Chat UI is the open code behind HuggingChat and supports only OpenAI-compatible endpoints, which makes it a developer's front end.
Chat that centres on your own documents
These are for asking questions of your own files. AnythingLLM comes as a desktop app or a Docker server for teams, and the Docker version adds multi-user use. Kotaemon offers document question answering with logins and shared collections; it's slower moving, with its last repository update in spring 2026, so check it's still active before relying on it.
The eight front ends, one line each
AnythingLLM
Best for: People who want to chat with their own documents, either as a desktop app or a multi-user Docker server for a team.
Big-AGI
Best for: Power users who want answers from several models side by side, to compare and combine, using their own keys.
Hugging Face Chat UI
Best for: Developers wanting the open code behind HuggingChat as a thin front end, knowing it supports only OpenAI-compatible endpoints.
Kotaemon
Best for: Small teams running a self-hosted question-answering interface over their own documents, with logins and private or shared file collections.
LibreChat
Best for: Organisations wanting one customisable multi-user web app that brings several AI providers, agents and tools together in a single interface.
LobeHub (LobeChat)
Best for: Teams wanting a chat and agent workspace they deploy with Docker Compose or one-click routes, and who won't distribute modified versions.
NextChat
Best for: Individuals or small groups wanting a light web chat client deployed through Vercel or Docker with their own model API keys.
Want AI doing the heavy lifting in your marketing?
I build the systems that handle the boring 80 percent, so you get your week back. Done properly, with the human kept in.
Open WebUI
Best for: Anyone running local models through Ollama or an OpenAI-compatible API who wants a self-hosted chat workspace that can run entirely offline.
Licence for each chat front end
Five of these are plain MIT or Apache 2.0, which let you use and modify the code freely. Two carry extra conditions. Open WebUI uses a BSD-style licence with a branding clause: you must keep its name and logo unless you have 50 or fewer end users in any 30 days or hold a separate licence, so it's source-available rather than plain open source. LobeHub's community licence is Apache 2.0 plus conditions: commercial use is allowed without changing the code, but building and distributing a derivative work needs a commercial licence.
| Tool | Licence | Good for |
|---|---|---|
| AnythingLLM | MIT | Chatting with your own documents, desktop or server |
| Big-AGI | MIT | Comparing and combining answers from several models |
| Hugging Face Chat UI | Apache 2.0 | A thin chat front end for OpenAI-compatible endpoints |
| Kotaemon | Apache 2.0 | Question answering over shared document collections |
| LibreChat | MIT | Several AI providers, agents and tools in one app |
| LobeHub (LobeChat) | Apache 2.0 plus conditions | An agent and chat workspace on your own server |
| NextChat | MIT | A light personal client deployed in minutes |
| Open WebUI | BSD-style with branding clause | An offline-capable team chat workspace for local models |
My advice for whoever is setting this up
Decide where the model runs before you compare interfaces, because that's where your privacy comes from. If you want nothing leaving the building, plan local hardware first and pick a front end that talks to local models well. If an API is fine, the choice is mostly about logins, sharing and how much you want to rebrand. And if you plan to put your own branding on it for a large team, read the licence before your logo goes anywhere.
Start small: one server, a handful of users, a single model, and a week of real questions. To choose the model side, our guide to open source language models is a good starting point.
Privacy depends on where the model runs
A self-hosted window keeps conversation history, accounts and uploaded files on your server, which is a real gain. It doesn't change what happens to a prompt once it's sent to a hosted model. Check that provider's data terms, set who can see whose chats, and decide how long histories are kept. Data protection rules vary by country, so check your obligations with your regulator rather than a vendor summary. If you're still deciding whether staff need a chat assistant or something that takes actions for them, our guide to AI agents for small businesses covers the other side.
Do you make a self-hosted AI chat front end? Tool makers can ask to be considered for a featured place on this list.
Want your tool on this list? Featured spots and reviews here are paid partnerships. Here's how to partner with me.
Frequently asked questions
Does a self-hosted AI chat interface include a model?
No. These tools are the chat window only. You connect a model yourself, either one running locally on your own hardware or a hosted model through an API key.
Is a self-hosted chat interface private?
It keeps chat history, accounts and files on your own server. But if you connect a hosted model through an API, your prompts still go to that provider. Full privacy needs a model running locally too.
Can I rebrand a self-hosted AI chat interface?
It depends on the licence. Plain MIT and Apache 2.0 tools allow it. One popular front end requires you to keep its name and logo above a small number of users unless you hold a separate licence.
What hardware do I need to run AI chat locally?
The chat window itself is light. The model is what needs power, usually a capable graphics card or a lot of memory for larger models. Small models can run on ordinary machines.
Can my team chat with our own documents privately?
Yes. Document-focused front ends let you upload files and ask questions of them, and with a local model nothing leaves your network. Check how each one handles user access to shared collections.
Do I need Docker to self-host an AI chat interface?
Most of these offer Docker as the main self-hosting route, and some offer a desktop app or one-click cloud deployment instead. Basic comfort with Docker makes set-up much easier.
