Kimi Launches Chrome Extension That Turns Web Actions Into Reusable Skills

The older WebBridge was designed as a bridge between a local Agent and Chrome or Edge extensions. When the Agent sent instructions, WebBridge could navigate pages and complete tasks in the user's real browser, then send the results back. In effect, it let an Agent use the user's browser. The new browser extension keeps that local Agent calling method, but it changes the entry point: users can click the Kimi icon in the toolbar, log in to their Kimi account, and talk to Kimi in the sidebar to operate the current webpage. It also retains the earlier browsing and information-scraping functions.

The main functional addition is Skill recording. Previously, when an Agent used WebBridge to operate a browser, it often needed repeated iterative tuning before it could complete a task efficiently. The new version can record a user's actions on webpages and organize them into a Skill. For example, if a user opens the same website every day, searches for the day's news, and exports data, the Kimi browser extension can save that set of webpage operations as a Skill and call it next time. If a task goes smoothly, Kimi can also save the session as a Skill. This makes the extension better suited to repetitive work, with processes that users can call repeatedly.

Kimi also warned that webpage redesigns and dynamically loaded pages may still cause operations to get stuck midway. When a complex page causes a failure, users can first take a screenshot to confirm the page state and then adjust instructions.

The browser extension follows a longer line of AI plugin development. The Manus team's earlier product Monica placed AI beside any webpage to help users ask questions and write emails. Doubao's plugin has emphasized video content filtering, translation of foreign-language webpages, and paper reading. That generation of AI plugins mainly addressed viewing webpages. In the using webpages stage, Manus later launched Browser Operator, which lets an Agent operate a user's local browser after authorization. Kimi WebBridge did something similar by controlling a real browser. The latest Kimi browser extension pushes one step further by giving users a way to save webpage processes for reuse.

QbitAI wrote that the extension's release shortly after Kimi Code Desktop is itself interesting. Kimi Code Desktop is a workbench for Agent projects, and its built-in browser follows a project, making it useful for checking the effect of webpage changes immediately. If something is wrong, users can directly select page elements, attach screenshots and comments, and ask the Agent to redo the work. Both products point to the browser from different sides: one operates inside a project, and the other operates outside in the user's daily applications. Together, the product idea is that an Agent on the desktop gets project files, a terminal, and a change record; when it needs to inspect a webpage it has made, it opens the built-in browser in the session; when it needs to enter websites the user uses daily, it works through the extension in existing Chrome or Edge. The article said the Agent's reach is extending longer and becoming more stable.

Editor's Summary

Kimi has released a Chrome extension that upgrades its earlier WebBridge with a sidebar interface and Skill recording for repeated webpage tasks. It follows Kimi Code Desktop, forming a desktop-plus-browser product setup for Agent work. The extension still carries limits around redesigned or dynamically loaded pages, which the company says may require screenshots and adjusted instructions.