Anthropic, the company behind the AI assistant Claude, has started testing a new Google Chrome extension. At the moment, the extension is limited to 1,000 Claude Max subscribers. This trial is a research preview, which means the company wants to see how people use it in real life before offering it to more users.
The Claude extension works like an extra helper inside the Chrome browser. After installation, it shows up on the right side of the screen as a panel with a chat box. People can use it while browsing any website.
With this tool, Claude can read the page, give short summaries, find details inside the text, and even answer questions about what is written there. It can also look at numbers or data included on a webpage and explain what they mean.
Right now, the extension cannot do automated actions, run tasks on its own, or control anything outside the page. Anthropic made this choice to avoid risks and keep the tool safe. Users are in control at every step.
For example, Claude asks for permission before reading a page. If an action could be risky, like sending or sharing personal information, the extension will ask for extra approval. Users also need to give access to each website separately before Claude can read it.
Anthropic says this slow and cautious launch is important for safety. One of the threats they are studying is called a “prompt injection attack.” This happens when hidden instructions inside a webpage trick an AI system into doing something harmful, like helping a phishing scam or spreading malware.
By running this trial with only 1,000 people on the paid Max tier, the company hopes to understand how the tool is used every day and what kinds of attacks it may face. Claude Max is not cheap subscriptions cost 100 dollars (around ₹8,700) or 200 dollars (around ₹17,500) per month but only people on this plan are allowed to join the test.
Currently, the extension is available only on the desktop version of Google Chrome. Anthropic has not said when it will be available more widely or whether it will be made for other browsers like Safari or Microsoft Edge.
A demo video shared on X (formerly Twitter) showed how the extension looks in use. The video showed users simply clicking the Claude icon, opening the sidebar, and then starting a conversation with the AI while reading a web page.
This new experiment comes after several recent updates to Claude. Earlier this year, Anthropic added the ability for Claude to end conversations if they turn harmful or abusive. The AI assistant can also now remember past chats and refer back to them later for more helpful answers.
The company has been rolling out new connections for Claude too, such as links with design and work tools like Notion, Canva, and Figma. At the same time, Anthropic introduced some limits on heavy coding use to prevent misuse.
The Chrome extension is one more step toward making Claude a tool people can use anywhere on the web, not only in a separate chat window. Right now, the main focus is on safety and testing, but the goal is clear. Anthropic wants Claude to become a trusted AI assistant that can make web browsing easier, faster, and more useful.