UWA (Universal Web Agent)

Local Autonomous Agent for Web AI Chats

A local automation tool that turns your Web AI chat into agent with direct access to your computer's terminal and file system. It connects to Chrome via the Chrome DevTools Protocol, monitors the chat, parses model tool calls, executes them on your system, and automatically pastes the results into out.json.

It effectively gives the AI persistent, real-world execution capabilities.

• Terminal Execution — Run any commands with full output capture
• File Operations — Safe read/patch/symbols (with ctags and Markdown support)
• Autonomous & Safe — Continuous monitoring with Double Patch Protection
• Auto-injection — Automatically pastes results back into the chat via $UWA_INPUT tag
• Chrome Integration — Automatically finds the active tab via DevTools Protocol

• grok.com
• work.trae.ai
• perplexity.ai
• app.kilo.ai
• duck.ai
• gemini.google.com
• chat.deepseek.com
• chat.qwen.ai
• claude.ai
• chatgpt.com
• alice.yandex.ru
• Most likely, it can work on many others without modifications

1. Launch Chrome with debugging enabled (--remote-debugging-port=9222)

2. Install dependencies:

pip install websockets

For Windows:

choco install universal-ctags -y

For Linux:

sudo snap install universal-ctags --classic

3. Run the agent:

python agent_executor.py

4. Start the session: Copy the content of agent.md and paste it into your Web AI chat. This initializes the communication protocol.

5. Execution Loop (Optional Auto-injection): When you want the agent to execute something, send your message to the AI. After the message is sent, type the $UWA_INPUT tag into the chat's input field. The agent will execute the requested tools and attempt to automatically inject the results back into that field.

Note: This is an experimental feature. It only works if you manually place the tag in the input field and may not be compatible with all AI chat interfaces yet.

• The agent only executes commands you explicitly request through tool calls
• Use with trusted AI models only

Disclaimer: This project is in the early MVP stage. You are fully responsible for the commands you allow the agent to execute and any consequences that may result.