Pinchtab Helper

What This Skill Does

The Pinchtab Helper skill empowers the OpenClaw AI agent to control a web browser efficiently using the PinchTab framework. Unlike traditional browser automation that relies on heavy, token-expensive screenshots, PinchTab utilizes accessibility references (accessibility trees) to identify elements. This results in a significantly lower token usage—roughly 800 tokens per page—making it between 5 to 13 times more cost-effective than visual-based agents. It enables your AI to navigate to URLs, fill out input fields, click buttons, and submit forms, providing a seamless bridge between local browser instances and AI decision-making. Whether you need to scrape data, manage web applications, or perform automated testing, this skill provides the necessary interface.

Installation

Before using the skill, ensure the PinchTab backend service is active. Install the tool via npm (npm install -g pinchtab) or use Docker (docker run -d --name pinchtab -p 127.0.0.1:9867:9867 pinchtab/pinchtab). Once installed, initiate the background process by running pinchtab &. Finally, install the OpenClaw skill via your terminal using the command: clawhub install openclaw/skills/skills/adminlove520/pinchtab-helper.

Use Cases

This skill is ideal for repetitive web tasks. Use it for automated research, such as scraping data from search engines or gathering pricing information. It is also highly effective for filling out recurring web forms, managing social media queues, or logging into web-based portals to extract status updates. Because it supports multi-instance operation, it can handle parallel browser tasks efficiently, saving developers and power users significant manual effort.

Example Prompts

1. "Open Google in a new browser instance and search for 'latest AI developments'."
2. "Go to the login page at [URL], enter my credentials into the fields, and click the submit button."
3. "Visit this page and give me a summary of the main navigation menu items found in the snapshot."

Tips & Limitations

To maximize effectiveness, always use a dedicated browser profile to keep your personal data separated from the automation tasks. Be mindful that while PinchTab is highly efficient, its success depends on the stability of the website's DOM accessibility properties. If a page uses complex, non-standard custom components, the accessibility refs may occasionally be elusive. For remote environments, ensure you configure the BRIDGE_TOKEN security variable. Always run the snap -i -c command frequently to refresh the AI’s understanding of the page state after every action taken.