LibreChat: Self-Hosted AI Chat Platform
danny-avila/LibreChat
LibreChat is a self-hosted AI chat platform that puts OpenAI, Anthropic, Google, AWS Bedrock, and other model providers behind one interface you run yourself. Reach for it if you want conversations and API keys on infrastructure you control and don't mind running Docker Compose. Skip it if you just want a chat window with zero setup or maintenance.
Unifying AI Conversations
LibreChat is an open-source, self-hosted web interface for chatting with AI models from OpenAI, Anthropic, Google, AWS Bedrock, Azure, and other providers through one login. It replaces juggling separate accounts and API dashboards with a single ChatGPT-style UI you deploy on your own servers. Model Context Protocol support lets it connect to external tools and data sources instead of being locked to one vendor's chat product.
Core Capabilities
- ✓Model switching across OpenAI, Anthropic (Claude), AWS Bedrock, Azure OpenAI, Google Vertex AI, and any OpenAI-compatible Custom Endpoint without a proxy
- ✓Code Interpreter API runs sandboxed Python, Node.js, Go, C/C++, Java, PHP, Rust, and Fortran, built on the open-source ClickHouse/code-interpreter
- ✓LibreChat Agents: no-code custom assistants with an Agent Marketplace, MCP server tools, file search, and code execution, plus reusable SKILL.md instruction bundles and Subagents for isolated child runs
- ✓Web search combines search providers, content scrapers, and result rerankers, with configurable Jina reranking
- ✓Code Artifacts generate React, HTML, and Mermaid diagrams directly inside the chat
- ✓Resumable Streams reconnect a dropped response automatically and sync the same conversation across multiple tabs and devices, scaling to Redis for horizontally scaled deployments
- ✓Multi-user auth via OAuth2, LDAP, or email login, with a browser-based Admin Panel for managing users, groups, roles, and config overrides without redeploying
- ✓Image generation and editing through GPT-Image-1, DALL-E (3/2), Stable Diffusion, Flux, or any MCP server
Getting Started with LibreChat
The README doesn't include a step-by-step command sequence in what's provided here—it points to Docker, reverse proxy, and other deployment options, and says the Admin Panel comes bundled with the Docker Compose stacks for one-command setup. For the actual install steps and prerequisites, check LibreChat's own documentation at librechat.ai/docs rather than guessing at commands.
Who Benefits from LibreChat?
LibreChat fits teams and individuals who already have API keys for multiple AI providers and want one place to use them instead of separate accounts for OpenAI, Anthropic, and Google. It suits developers comfortable running Docker and maintaining their own server, and organizations that need multi-user authentication (OAuth2, LDAP, email login) with an admin panel to manage who can use what. It's a poor fit for someone who just wants a single hosted chatbot with no infrastructure to manage.
Strengths
- ✓One interface for a long list of providers—including AWS Bedrock, Vertex AI, and OpenAI's Responses API—so switching models mid-conversation doesn't mean switching apps
- ✓MCP support and Agents give it real extensibility beyond a chat box: tools, file search, and code execution attach to an agent rather than living in a separate plugin system
- ✓Resumable Streams and multi-tab sync are the kind of detail most self-hosted chat UIs skip—useful if your team leaves tabs open all day
- ✓MIT license with 41,710 GitHub stars and 8,609 forks behind it, which at least means a lot of other people are running the same code
Considerations for Self-Hosting
- △The README doesn't give a single install command or prerequisite list here—expect to read the separate docs site before you get a working Docker Compose stack
- △Every provider integration (Bedrock, Vertex AI, Azure, Custom Endpoints) is a config surface you now own; wiring up more than one or two providers means more environment variables to keep straight, not fewer
- △Self-hosting means you're responsible for the update cadence yourself—the README specifically flags breaking changes in the changelog, which implies upgrades aren't always drop-in
Comparing AI Chat Platforms
Frequently Asked Questions
LibreChat connects to OpenAI, Anthropic (Claude), AWS Bedrock, Azure OpenAI, Google Vertex AI, and Local & Remote providers including Ollama, Mistral AI, OpenRouter, and Deepseek, plus any OpenAI-compatible API through Custom Endpoints.
LibreChat itself is free and open-source under the MIT license—you self-host it, so the only ongoing cost is your own server plus whatever API usage fees the AI providers you connect charge separately.
LibreChat supports multi-user, secure authentication through OAuth2, LDAP, and email login, with an Admin Panel for managing users, groups, roles, and permissions from the browser.
LibreChat generates images with GPT-Image-1, DALL-E (3/2), Stable Diffusion, or Flux, and it can analyze uploaded images with vision-capable models like Claude 3, GPT-4o, and Gemini.
LibreChat Agents are no-code custom assistants you build inside the app, with an Agent Marketplace for discovering community-built ones and support for MCP server tools, file search, and code execution.
LibreChat runs on infrastructure you control instead of sending conversations to a third-party chat product, and its Code Interpreter runs code in an isolated, sandboxed environment kept separate from your own systems.
The problem it solves
Anyone paying for OpenAI, Anthropic, and Google APIs separately ends up with conversations scattered across three different vendor UIs, each with its own login, history, and file-upload quirks—and no way to keep that data off those vendors' servers. LibreChat's answer is a single self-hosted interface that talks to all of them through Custom Endpoints and native provider integrations. Switching models is a dropdown, not a context switch.
Best use cases
- •Running one internal chat UI for a team that already pays for OpenAI, Anthropic, and Google APIs separately, so people stop copy-pasting between three tabs
- •Building a no-code agent that calls MCP tools and file search for a narrow internal workflow, shared with a specific group instead of company-wide
- •Self-hosting AI chat for a company that can't send prompts to a third-party chat product for compliance reasons
- •Prototyping with Code Interpreter to test code in Python, Go, or Rust against a model's output before shipping it anywhere else