LibreChat: How to Self-Host It with Docker on Windows and Mac

By Shah Rukh, software developer · October 2, 2026 · 7 min read

LibreChat is a free, open-source chat app that looks like ChatGPT but runs on your own computer and can talk to many AI providers from one screen. This guide shows how to run it with Docker on a Windows PC or a Mac, what it really costs, and what to check before you start.

Read this first. LibreChat is only the chat interface. It does not include any AI model. The answers come from a provider such as OpenAI, Anthropic or Google, and their API keys are normally paid, charged by usage. The only way to avoid that is a free tier from some provider or a local model on your own machine. LibreChat is an independent project and is not made by any of those companies.

What LibreChat is

The project describes itself as “a self-hosted AI chat platform that unifies all major AI providers in a single, privacy-focused interface”. Self-hosted means you run the software yourself instead of using someone’s website.

The version in the repository is v0.8.8, and it is released under the MIT licence, which allows free use, including commercial use. The README lists support for Anthropic (Claude), OpenAI, Azure OpenAI, Google, Vertex AI and AWS Bedrock, plus “custom endpoints” for any OpenAI-compatible service such as Ollama, Groq, Mistral, OpenRouter and Deepseek.

Beyond plain chat, the project lists agents (custom assistants you build without code), file uploads, web search, image generation, speech, conversation import and export, and multi-user login. Many of these need extra keys or setup.

With Docker, LibreChat is not one program. The default docker-compose.yml file starts six containers (small isolated programs):

What you need

• A Windows PC or a Mac with Docker Desktop installed and running. Docker Desktop is free for personal use.
• Git, to download the project.
• A text editor such as Notepad or TextEdit.
• At least one AI provider key, or a local model. This is where the cost is. Check the provider’s own pricing page, because the repository does not list prices.

The repository does not state minimum RAM or disk space. Six containers and several downloaded images are a real load, so an older laptop with little memory may struggle. You do not need a graphics card unless you also run a local model.

You do not need Node.js for the Docker method. The project’s .nvmrc file names Node 24.16.0, but that is only for people who run it without Docker.

Precautions before you install

1. Use the real repository. Clone only from github.com/LibreChat-AI/LibreChat. Popular projects attract fake copies.
2. The first account becomes the admin. In the code, the first user who registers gets the admin role, and ALLOW_REGISTRATION=true is the default. Create your own account straight after the first start.
3. The ports are open to your network. The compose file publishes ports 3080 and 3000 without limiting them to your own computer. On hostel, office or café Wi-Fi, other people may be able to open your sign-up page. Use a trusted network, keep your firewall on, and turn registration off once your account exists (see below).
4. Your chats still go to the AI provider. Self-hosting keeps your chat history on your own disk, but every message and uploaded file you send to a cloud model goes to that company. Don’t paste passwords or client data.
5. Protect the .env file. It holds your secrets and any keys you add. Never share it or upload it to GitHub.
6. Don’t expose the databases. The bundled MongoDB runs with no password (--noauth) and the file database uses the sample password mypassword. That is acceptable only because their ports are not published. Don’t add port mappings for them.
7. The default image is a development build. The compose file pulls librechat-dev:latest. It gets the newest changes, and sometimes the newest bugs.

How to install LibreChat with Docker on Windows

The README does not print install steps. It points to the project’s documentation site. The steps below follow the compose file and .env.example in the repository.

1. Start Docker Desktop and wait until it says it is running.
2. Open PowerShell from the Start menu and download the project:

   git clone https://github.com/LibreChat-AI/LibreChat.git
   cd LibreChat

3. Make your settings file by copying the example:

   copy .env.example .env

4. Open it in Notepad:

   notepad .env

5. Find the line ADMIN_PANEL_SESSION_SECRET= and put a random value of at least 32 characters after the equals sign. The compose file notes that the admin panel refuses to start without it. Save and close.
6. Start everything:

   docker compose up -d

The first start downloads all the images, so it can take a while on a slow connection.

How to install LibreChat with Docker on Mac

The steps are the same. Only two commands differ.

1. Start Docker Desktop.
2. Open Terminal (Cmd + Space, type Terminal) and run:

   git clone https://github.com/LibreChat-AI/LibreChat.git
   cd LibreChat
   cp .env.example .env

3. Create the random secret with the command the project suggests, and copy the output:

   openssl rand -hex 32

4. Open the settings file, paste the value after ADMIN_PANEL_SESSION_SECRET= and save:

   open -e .env

5. Start everything:

   docker compose up -d

If you see file-permission errors, note that .env.example has two commented lines, UID=1000 and GID=1000, which the compose file uses for the user the containers run as. The repository does not explain when to change them, so check the project’s documentation.

First use: create your account and add a key

1. Open http://localhost:3080 in your browser.
2. Click the sign-up link and register with an email and password. The account is stored only in your local database, and it becomes the admin.
3. Pick a provider in the model menu. In .env.example, OPENAI_API_KEY, ANTHROPIC_API_KEY and GOOGLE_KEY are set to user_provided, so the app expects each user to enter their own key in the interface.
4. Send a short test message.

If you would rather set one key for everyone, replace user_provided with the key in .env and restart. The admin panel is at http://localhost:3000.

Useful settings in the .env file

After changing .env, run docker compose down and then docker compose up -d.

Other providers and local models: services such as Groq, OpenRouter or Ollama are added in a second file, librechat.yaml. The repository includes librechat.example.yaml as a starting point, and docker-compose.override.yml.example shows how to load it: you save a copy as docker-compose.override.yaml and uncomment the “USE LIBRECHAT CONFIG FILE” section. The same file shows how to switch to the numbered release image, registry.librechat.ai/librechat-ai/librechat:latest. The project asks you not to edit docker-compose.yml directly.

How to stop, update and remove LibreChat

To stop it, run docker compose down inside the LibreChat folder. Your chats stay in the data-node folder, and uploads in uploads and images.

To update, the README asks you to read the changelog for breaking changes first. The repository also has an UPGRADING.md file, which says to back up MongoDB before database migrations. Copy the data-node folder while the app is stopped. Then:

docker compose down
git pull origin main
docker compose pull
docker compose up -d

The project also ships a script, npm run update:docker, but it needs Node.js installed and rebuilds the image on your machine.

The repository does not document an uninstall. Stopping the containers and deleting the LibreChat folder removes the app and your chat history. Unused images can be cleaned up from Docker Desktop.

Common problems

To see what went wrong, run docker compose logs api, or look in the logs folder. You can report bugs on the project’s GitHub issues page.

This guide is based on the project’s repository as it was on October 2, 2026. The project changes often, so if a step looks different, follow the official documentation linked from the README. ToolsCloset is not connected to the project or to any AI provider named here.

More AI tips

Related tools