Guide · Open source
How to use LibreChat: setup, providers and first chat
Updated · written with AI from the project's own documentation and checked against our catalogue data
LibreChat is a free, open-source web app that looks like ChatGPT but lets you connect to many AI providers, including OpenAI, Anthropic (Claude), Google, Azure, and local models through compatible APIs. You run it on your own server, so conversations and accounts stay under your control. This guide explains what you need before installing, how the main pieces fit together, and how to get your first chat running.
LibreChatLibreChat-AI/LibreChat · ★ 46.6k · MIT Install options & downloads Quick start
- 1Check prerequisites. Confirm you have Git and Docker on the machine that will host LibreChat.
- 2Clone the repository. Run git clone https://github.com/LibreChat-AI/LibreChat to copy the project.
- 3Follow the README setup. Work through the build and deploy steps in the README and the docs at docs.librechat.ai.
- 4Add your API keys. Configure keys for the AI providers you want to use, such as OpenAI or Anthropic.
- 5Start a chat. Open the web app, pick a model or endpoint, and send your first message.
- 6Add users and presets. Set up accounts for your team and save presets for common configurations.
What you need before you install LibreChat
LibreChat is a self-hosted application, so you need a server or computer to run it on. The project's install method is clone-and-build, and the prerequisites listed for it are Git and Docker. The README's deploy options point toward Docker-based setup, though this has not been verified against the full docs here.
You will also need API keys for any AI providers you want to use, because setup involves configuring those keys on your server.
- Git, to copy the repository
- Docker, the likely setup route based on the README's deploy options
- API keys for the providers you plan to connect
Clone the LibreChat repository
Start by copying the project from GitHub with Git. The repository lives under the LibreChat-AI organisation. After cloning, follow the setup instructions in the project's README and documentation at docs.librechat.ai for the remaining configuration steps.
git clone https://github.com/LibreChat-AI/LibreChatConnect AI providers and models
LibreChat supports a wide range of model sources. The README lists Anthropic (Claude), AWS Bedrock, OpenAI, Azure OpenAI, Google, Vertex AI, and the OpenAI Responses API. You can also reach local or remote providers such as Ollama, Mistral AI, Groq, OpenRouter, and DeepSeek.
For any service that speaks the OpenAI API format, the README describes Custom Endpoints, which let you use it without a proxy. Once providers are configured, you can switch between endpoints and presets in the middle of a chat.
- Built-in providers: Anthropic, AWS Bedrock, OpenAI, Azure OpenAI, Google, Vertex AI
- Local and remote: Ollama, Groq, Mistral AI, OpenRouter, Perplexity, DeepSeek and others
- Custom Endpoints: any OpenAI-compatible API
Use presets, branching and message tools
Presets let you create, save and share a configuration of endpoint and settings. Within a conversation you can edit, resubmit and continue messages, which creates conversation branches. You can also fork messages and conversations to control context more precisely, and compact long conversations on demand while keeping recent context.
- Save and share presets with specific users and groups
- Edit or resubmit a message to branch the conversation
- Fork messages or conversations for more control over context
Work with agents, MCP tools and code
Agents are no-code custom assistants. An agent can use MCP servers, tools, file search and code execution. Skills are reusable SKILL.md instruction bundles. Subagents delegate focused work to isolated child runs.
The Code Interpreter API runs code in sandboxed environments, including Python, Node.js, Go and Rust, and supports uploading and downloading files. Note that the README marks attached code workspaces as highly experimental, so treat them with caution.
- Agents: no-code assistants that can use MCP servers and tools
- Skills: SKILL.md instruction bundles for agent workflows
- Code Interpreter: sandboxed execution for many languages
Set up multi-user access
LibreChat includes secure multi-user authentication, which suits teams that need shared login and role-based access. The README also mentions sharing agents and prompts with specific users and groups. Check the documentation for the exact auth configuration options for your deployment.
What people use it for
- Give a team one shared chat interface across several AI providers
- Run a private ChatGPT-style assistant on your own server
- Connect a custom OpenAI-compatible endpoint or a local model
- Build no-code agents that use MCP tools and file search
Frequently asked questions
Is LibreChat free?
LibreChat is an open-source project released under the MIT licence, and the source describes it as a free web app. You still pay for any AI provider API usage you connect, and you cover your own hosting costs.
What do I need to install LibreChat?
The install method is clone-and-build with Git and Docker as the listed prerequisites. You also need API keys for the AI providers you want to use. Setup involves configuring a server, so it suits people with some hosting experience.
Can I use LibreChat with local models?
Yes. The README says LibreChat is compatible with local and remote providers such as Ollama, Apple MLX and koboldcpp, and with any OpenAI-compatible API through Custom Endpoints.
Is LibreChat better than Open WebUI?
Both are self-hosted, ChatGPT-style interfaces. LibreChat emphasises agents, MCP support and multi-provider access, while Open WebUI, LobeChat, Chatbot UI and AnythingLLM are comparable projects you can evaluate against your own needs.
Do I need a server to use LibreChat?
Yes, LibreChat is designed to run on your own server. If you only want to chat with AI and do not want to run a server, a hosted service is simpler.
Is LibreChat safe for multiple users?
The README lists secure multi-user authentication and role-based access as part of the project, which suits teams. Review the documentation to configure authentication correctly for your deployment.
New agent skills, repos and AI tools — every week
The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.
Our verdict
LibreChat suits teams and developers who want one self-hosted, ChatGPT-style interface across many AI providers, with multi-user login and agents. It is a poor fit if you have no server or Docker experience, or if you only want a hosted chatbot. Features the README marks as experimental, such as attached code workspaces, should be tested before you rely on them.
Install options & downloads