Build Your Own AI Chat Hub with LibreChat: One-Click Model Switching

What is LibreChat

LibreChat is an MIT‑licensed, self‑hosted AI aggregation platform that unifies access to OpenAI, Anthropic, Google, Azure, AWS, Groq, Mistral, DeepSeek, Ollama and other compatible models behind a single web UI while keeping all data on the user’s own infrastructure.

Key Features

Unified Model UI

The interface mirrors ChatGPT with a conversation list, chat window and input box. A top‑dropdown lets the user switch among models such as GPT‑5, Claude Sonnet 4.6, Gemini 2.5, DeepSeek V3, or a locally‑run Ollama Llama 3 model. Model changes can be made mid‑conversation, allowing the same thread to continue with a different backend.

Agent and MCP Tools

LibreChat ships a drag‑and‑drop Agent framework that requires no code to create custom assistants. Agents can invoke MCP servers, execute code, search files, or call external APIs. Community‑shared Agent configurations can be imported and edited; the SKILL.md mechanism enables scripted workflows, sub‑agents, and background execution.

Code Interpreter

A built‑in sandboxed interpreter supports Python, Node.js, Go, Java, C/C++, PHP, Rust and Fortran. Users upload files (e.g., CSV), the model generates a Python script, the script runs in the sandbox, and results are returned as tables or charts directly in the chat. The interpreter is powered by ClickHouse’s open‑source code‑interpreter project.

Artifacts and Image Generation

Models can emit React components, HTML pages or Mermaid diagrams as “Artifacts,” displayed side‑by‑side with the chat. Image generation integrates DALL‑E 3, Stable Diffusion, Flux and GPT‑Image‑1, supporting text‑to‑image, image‑to‑image and image‑editing workflows, and can connect to any MCP server offering image generation.

Multimodal and File Handling

Uploaded images, PDFs or videos can be processed by visual models such as Claude 3, GPT‑4o or Gemini. PDFs are parsed so the model can answer questions based on document content. A separate rag_api service vectorizes uploaded documents into a pgvector store for retrieval‑augmented generation.

Conversation Management and Search

Conversation tagging, folder organization and full‑text search are powered by MeiliSearch, enabling instant lookup of messages from months ago. The Fork feature creates a new branch of a conversation, preserving the original while testing alternative models or prompts. A reconnection mechanism resumes streaming output after network interruptions and synchronizes sessions across devices.

Quick Start

Deploy LibreChat with Docker Compose on a Linux host with at least 4 GB RAM.

git clone https://github.com/danny-avila/LibreChat.git
cd LibreChat
cp .env.example .env
# edit .env to add API keys for OpenAI, Anthropic, Google, etc.
docker-compose up -d

The compose file pulls MongoDB, RediSearch and the LibreChat service containers. After the containers start, open http://<host>:3080 in a browser, register the first account (automatically an admin), and use the built‑in admin panel to manage users, roles, API keys and model parameters without restarting the service.

https://github.com/danny-avila/LibreChat