LibreChat

LibreChat is a free, MIT-licensed, self-hosted AI chat platform that brings together language models from every major provider into a single unified interface. Originally created by a single developer, it has grown into a community project with over 413 contributors and, according to the project's own homepage, 44,300+ GitHub stars and 53.1 million Docker pulls. The project is joining ClickHouse to power what they describe as the open-source Agentic Data Stack.

What It Is

LibreChat is a self-hosted web application—not a native desktop app—that acts as a universal front-end for AI models. Users deploy it via Docker, npm, or cloud hosting, then connect it to any combination of AI providers without vendor lock-in. It is released under the MIT License, meaning it is free to use, modify, and distribute.

Multi-Provider Model Access

One of LibreChat's defining features is its breadth of supported AI backends:

• Commercial APIs: OpenAI (including GPT-5 and Responses API), Anthropic (Claude), Google (Gemini, Vertex AI), Azure OpenAI, AWS Bedrock
• Custom & local endpoints: Ollama, Groq, Mistral AI, Cohere, Deepseek, Qwen, OpenRouter, Perplexity, koboldcpp, together.ai, Apple MLX, and more
• Any OpenAI-compatible API can be added as a custom endpoint without a proxy

Users can switch between models and endpoints mid-conversation, save presets, and share configurations across teams.

Agents, MCP, and Extensibility

LibreChat includes a no-code agent builder that supports the Model Context Protocol (MCP), custom skills, subagents, and an agent marketplace for community-built helpers. Key agent capabilities include:

• Skills: Reusable SKILL.md instruction bundles for manual, automatic, or always-on workflows
• Subagents: Delegate focused work to isolated child agent runs with their own context windows
• Agent Plugins: Bundle deployment skills and MCP servers into startup-loaded packages
• Attached Code Workspaces: Let agents inspect, search, edit, and run commands in managed workspaces (marked highly experimental)
• Agent Management API: Automate agent, file, and skill management with deployment-bound OIDC clients

Code Interpreter and Artifacts

LibreChat ships a sandboxed Code Interpreter API—powered by the open-source ClickHouse/code-interpreter—that executes Python, Node.js (JS/TS), Go, C/C++, Java, PHP, Rust, and Fortran in fully isolated environments. A Generative UI / Artifacts feature lets models produce interactive React components, HTML pages, and Mermaid diagrams directly in the chat window, with fullscreen preview and SVG/PNG export.

Enterprise and Self-Hosting Features

LibreChat is designed for organizations that need control over their AI infrastructure:

• Authentication: OAuth2, SAML, LDAP, and two-factor auth for enterprise SSO
• Admin Panel: Browser-based UI to manage users, groups, roles, and live configuration overrides without redeploying
• Observability: OpenTelemetry log and trace export; Langfuse integration for agent and model insights
• Resumable Streams: AI responses automatically reconnect if a connection drops; multi-tab and multi-device sync supported
• Multilingual UI: Interface available in 30+ languages with community-driven translations
• S3/CloudFront support for stable media links and edge delivery

Update: v0.8.8-rc3

The latest release candidate (v0.8.8-rc3) introduces an Agent Management API in beta, attached code workspaces (highly experimental), manual context compaction, a unified attachments system, GPT-6 Astra model support via the Responses API, OpenTelemetry observability export, and a range of reliability and security improvements including stronger tenant isolation, Redis liveness detection, and MCP OAuth session hardening. The full changelog is published at librechat.ai/changelog.