HuggingChat UI
Self-hosted AI chat interface under Apache 2.0. Connect it to Ollama, llama.cpp or cloud APIs, with optional model routing and MCP tools.
HuggingChat UI is the open-source chat application behind Hugging Face's hosted HuggingChat. You can run it on your own computer or server and connect it to a local LLM backend or a cloud provider. It's for people and teams who want a browser-based ChatGPT alternative with control over the chat service and where its data lives. The code uses the Apache 2.0 license.
The interface works with OpenAI-compatible APIs, including Ollama, llama.cpp server, OpenRouter and Hugging Face Inference Providers. Model choice depends on the connected backend. The UI doesn't run models itself: a local backend handles inference on your hardware, while a cloud connection sends requests to that provider. Hugging Face's public HuggingChat is a hosted service, separate from your own deployment.
You can choose a model directly or use the optional Omni router to assign messages to different models. It distinguishes image requests, requests that need tools and ordinary chat. Routing uses local server logic rather than a separate model-selection service.
MCP support lets models call tools from connected servers and use their results in the conversation. The chat displays tool parameters, progress and results or errors. Models need function-calling support for this feature.
Chat history, user records, settings and files live in MongoDB, which can run locally or through a managed service such as MongoDB Atlas. A Docker image bundles the app with MongoDB. You can also customize the app's name, logos and favicons.
