Saltar al contenido principal

Libre WebUI

Make whatever comes next.

Run Libre WebUI on infrastructure you control. Give supported models governed tools, ground answers in cited documents, work with a team, speak in turn-based voice mode, create and edit media, and carry the same workspace from a laptop to a multi-replica deployment.

One command, no Libre WebUI vendor account

npx libre-webui@latest

Open the workspace at http://localhost:8080. Add Ollama for local models or connect a supported provider when you choose.

What do you want to do?

01

Start on your own infrastructure

Run one command, use a local Ollama model or a supported provider, and keep deployment choices in your hands.

Install Libre WebUI
02

Let models act safely

Use built-in, OpenAPI, or HTTP MCP tools with live call state, per-user access, and approval before side effects.

Configure Chat Tools
03

Ground answers in your knowledge

Search PDF, Office, Markdown, code, and data files with hybrid retrieval and trace answers to their source locations.

Use cited Document Chat
04

Work with a team

Bring people and models into channels, share resources through live grants, and coordinate with notifications and calendars.

Explore team collaboration
05

Speak and create media

Hold turn-based voice conversations, generate speech and video, and edit or inpaint images in one per-user gallery.

Start with Voice Mode
06

Operate it in production

Add budgets, evaluations, structured telemetry, recovery gates, and a release-tested multi-replica team profile.

Plan the platform

Everything is one tab away.

Compose an assistant

Assistant Profiles bind a model, prompt, tools, skills, knowledge collections, and voice while rechecking the current user's access whenever the profile runs.

Keep durable working material

Chats, isolated Work projects, and Markdown Notes persist independently. Notes add revision restore, attachments, sharing, export, and reversible AI-assisted edits.

Search instead of hunting

Press Cmd/Ctrl + K to reach an action or search authorized chat, note, and document content. Search runs over decrypted candidates in memory without a plaintext index on disk.

Find the path that matches your work.

Start and choose models

Install the workspace, size local hardware, and connect only the model routes you intend to use.

Agents and reusable assistants

Move from a governed tool call to a durable coding workspace or a reusable assistant profile.

Knowledge, notes, and schedules

Keep source-grounded research, durable Markdown, live web results, calendars, and scheduled AI work together.

Teams, voice, and media

Collaborate through live grants and channels, then add governed speech and per-user media workflows.