🖥️ ClawTear — a chat UI for the Ollama models already on your own computer

[NOBITHEMADNESS]

ClawTear — a chat window for the AI models already on your own computer

A member-built interface for Ollama — a full chat app that never leaves your machine.

What is actually in it

• Deep Think — reasoning models stream their step-by-step thinking, with a timer
• attach a code or document file, and read any answer aloud
• branch a conversation from any message; export chats as Markdown or JSON
• temporary chats, pinning, rename and search across your whole history
• tokyo-night among four themes, with temperature, Top-P and Top-K in reach
• no npm install and no build step — it runs on Node’s own built-ins

Download — claw-tear.zip (1.8 MB)

Three steps to running it

1. Unzip the folder
2. Double-click Start Local AI.bat
3. Your browser opens localhost:3000

Every line is readable. What you download is source — src/App.tsx, server.js, package.json — so you can read it first and run it second.

What is in the zip

• the whole React + TypeScript app as source — src/App.tsx and src/components/MarkdownRenderer.tsx
• server.js, the local server that proxies Ollama’s own /api/chat, /api/tags and /api/version
• Start Local AI.bat, run.bat, diagnose.bat, start.sh — Windows and Linux launchers
• README.md with the model list and the --port, --host and --ollama flags
• a prebuilt public/assets bundle, so it opens before anything is built

What it needs

• Node.js — the one requirement the submission names, and no npm install follows it
• Ollama with a model pulled; the screenshots show llama3.2:3b, and qwen2.5:1.5b is lighter still
• Deep Think mode wants a reasoning model — deepseek-r1:1.5b, or the 7b when the machine is roomy
• port 3000 free on your machine

Built by a member here — the whole app, free, source included.