HuggingChat

What it does

HuggingChat is a chat app that takes typed messages (and images, where a model supports them) and returns text answers from open source AI models. A model called Omni can pick which model answers each request, or you can choose a specific open model yourself. A newer experimental mode, ML Intern, handles machine-learning tasks on the Hugging Face Hub: it reads papers, models and datasets, writes scripts, and runs them as Hugging Face Jobs and sandboxes.

People use it as a general chat assistant, or switch to ML Intern to reproduce papers, finetune and evaluate models, build datasets, or deploy demos as Spaces, within a compute budget set per conversation. It can also call tools from Model Context Protocol servers. What sets it apart is that the code behind it, Chat UI, is public under Apache-2.0, so you can run your own version against any OpenAI-compatible endpoint.

It suits people who want to compare open models in one interface, ML practitioners working on the Hugging Face Hub, and developers or teams who want to self-host a chat front end. Self-hosting needs Node.js and an OpenAI-compatible API key or endpoint; MongoDB is optional locally. Generated content may be inaccurate or false, and ML Intern is labelled experimental, with billing setup required for its compute.

What it can do

• Chat with open source AI models
• Omni automatic model routing based on request type (image, tools, default)
• Choose a specific open model to chat with directly
• Call tools from Model Context Protocol servers
• ML Intern mode: read papers, models and datasets, write scripts, run Hugging Face Jobs
• Self-host the open source Chat UI with Docker or npm

Works with

• Model Context Protocol (MCP) servers
• Hugging Face Inference Providers router
• Slack (ML Intern bot)
• OpenAI-compatible APIs (llama.cpp, Ollama, OpenRouter, Poe)

Example workflows

Reproduce a paper with ML Intern

Input: a paper and a compute budget. Steps: open ML Intern mode, describe the task, let it read the paper, write scripts and run them as Hugging Face Jobs. Output: scripts, run results and possibly a demo Space.

Self-host a chat front end

Input: an OpenAI-compatible endpoint and API key. Steps: clone chat-ui, set OPENAI_BASE_URL and OPENAI_API_KEY in .env.local, run npm install and npm run dev. Output: a local chat interface at localhost:5173.

Before you choose

• Omni routing in self-hosted Chat UI is heuristic and needs a routes policy JSON you create yourself
• ML Intern is experimental and requires account and billing setup with a per-conversation compute budget
• Chat UI supports only OpenAI-compatible APIs; legacy provider-specific integrations were removed

Listed under

• Productivity
• General assistants
• AI models
• Assistant
• Chat
• Chatbot
• Developers
• General purpose

Get started

Use the Start chatting button on the hosted page

1. Open huggingface.co/chat and select Start chatting
2. Choose Omni or a specific model; use ML Intern mode via ?mode=ml-intern
3. For self-hosting, create .env.local with OPENAI_BASE_URL and OPENAI_API_KEY, then run npm run dev

Check the hosted app for current access and usage limits. The Apache-2.0 Chat UI can be self-hosted, with model endpoints and compute paid separately.

We couldn’t find published prices for HuggingChat. See their pricing page ↗

Sources for this overview

Product pages, documentation, and pricing checked 28 Oct 2026. Each source supports part of the overview; availability and prices can change.

• HuggingChat - Chat with AI models ↗huggingface.co
• GitHub - huggingface/chat-ui: The open source codebase powering HuggingChat · GitHub ↗github.com

Company website

huggingface.co: product, pricing and policy pages

YC directory mirror

Batch and company name

Web search

Confirms the official site

Social accounts

Accounts the company links, and its posts on X

Human review

Checked by a person

knowledge_base

Last checked 28 Oct 2026. Every field keeps its source and date; unknown means we couldn’t confirm it.