Open WebUI

Open WebUI is a self-hosted AI chat platform that runs fully offline and works with both local and cloud models. It connects to local Ollama models and to any OpenAI-compatible API, such as LM Studio, GroqCloud, Mistral, OpenRouter, or vLLM, so one interface can mix providers. You can wrap any base model with custom instructions, tools, and knowledge to build specialized agents, and talk to several models at once in one conversation.

Retrieval-augmented generation (RAG) runs locally on a choice of 9 vector databases and several content-extraction engines (Tika, Docling, Document Intelligence, Mistral OCR, PaddleOCR-VL), with hybrid BM25 and vector search, reranking, and a full-context mode. Web search pulls results from dozens of providers into the conversation, and the model can fetch web pages on its own.

It extends through Filters, Actions, Pipes, Tools, and Skills, and connects external services through MCP, MCPO, and OpenAPI tool servers. With Open Terminal, agents get a terminal and file system to run scripts, analyze data, and produce files in the chat. Beyond chat, it includes Notes, shared Channels where people and models work in one timeline, persistent memory across conversations, calendars that models manage through function calling, scheduled automations, voice and video calls, and image generation and editing.

For teams, administrators get role-based access control with user groups, LDAP and Active Directory, SSO, SCIM 2.0 provisioning, usage and cost dashboards, and a model evaluation arena with ELO leaderboards. It stores data in SQLite or PostgreSQL, files locally or in S3, Google Cloud Storage, or Azure Blob Storage, emits OpenTelemetry traces, and scales horizontally with Redis-backed sessions. It runs as a Python package, in Docker (including images bundled with Ollama or CUDA), or on Kubernetes, and a native desktop app is available for macOS, Windows, and Linux.