How to Use LibreChat

LibreChat is an open-source ChatGPT clone that talks to every major model provider from a single UI.

LibreChat started as a faithful ChatGPT reimplementation and grew into a multi-provider chat hub. One interface fronts OpenAI, Anthropic, Google, Azure, AWS Bedrock, local Ollama, and the OpenAI Assistants API. It adds plugins, multimodal inputs, presets, and per-user API keys on top.

What it is

A Node.js + React monorepo backed by MongoDB for chat history and Meilisearch for conversation search. Auth supports email, Google, GitHub, Discord, Facebook, Apple, and OpenID. Models are configured via librechat.yaml, which is where you declare endpoints, custom providers, and rate limits.

Install / set up

# docker compose
git clone https://github.com/danny-avila/LibreChat
cd LibreChat
cp .env.example .env
cp librechat.example.yaml librechat.yaml
docker compose up -d

First run

Open http://localhost:3080, register an account (the first user can be elevated to admin via the CLI), add an API key in the top-right menu, and start chatting. Switch providers from the model dropdown mid-conversation to compare outputs side by side.

$ npm run create-user admin@example.com Admin password123
User created successfully
$ npm run ban-user someone@example.com

Everyday workflows

• Save a Preset with a provider, model, system prompt, and temperature so teammates start from the same baseline.
• Enable Assistants to upload files and run OpenAI’s code interpreter without leaving LibreChat.
• Wire in plugins (web search, DALL·E, Zapier, custom actions) to extend chat into tool-calling agents.

Gotchas and tips

librechat.yaml is the source of truth for custom endpoints, but it’s read at boot — changes require a container restart. Validate your YAML before restarting, because a typo will crash the API container and silently fall back to the last good config only if you’re lucky.

MongoDB grows fast if users never delete conversations. Set up a retention policy or a cron that archives old threads, and put Meilisearch on a persistent volume or you’ll lose full-text search indexes on every redeploy.

Who it’s for

Anyone who wants a polished, self-hosted ChatGPT with multi-provider support and real auth. Families sharing a plan, small teams wanting unified billing through API keys, and tinkerers who like the ChatGPT UI but want to own the data.