Skip to content
OLLAMA · DIRECT FROM YOUR BROWSER

A local model needs
a usable interface.

Connect pui.ai to Ollama on this computer and keep the product features of a serious chat workspace. No pui.ai inference proxy sits between the browser and your local model.

Open the Ollama connectionSetup guide
Direct loopback requestsModel discoveryOptional Bearer tokenConnection pause switch
THE WORKSPACE LAYER

More than a local chat window.

01

Named local runners

Rename the Ollama connection for this workspace, change its endpoint, add an optional token and keep the immutable connection identity safe.

02

Clear diagnostics

Check the endpoint, discover models and optionally run a tiny generation test. Every stage reports its own result and duration.

03

Files and Vision

Attach bounded documents and processed images when the discovered model has the capabilities required for that request.

04

Organized local work

Use folders, colors, projects, prompt libraries, agents, branches, queues and complete local backups around Ollama chats.

CONNECTION SETUP

Three explicit steps.

  1. 1

    Run Ollama

    Start Ollama and verify its OpenAI-compatible API is available at http://127.0.0.1:11434/v1.

  2. 2

    Allow this secure origin

    Configure Ollama to accept https://pui.ai, for example with OLLAMA_ORIGINS=https://pui.ai ollama serve, and approve the browser's Local Network Access prompt.

  3. 3

    Check and discover

    Choose Ollama in Models & API keys, select “Check endpoint & discover models,” import the returned catalog and save the connection.

HTTPS and local HTTP

Modern Chromium browsers can grant a secure site explicit access to loopback. Ollama still controls CORS, and other browsers may apply different local-network policies.

Ollama is already running?

Connect it, discover models and keep every other provider optional.

Open the workspace