Moonshot AI's Kimi Browser Extension Turns Web Actions Into Reusable Skills
The AMW Read
Incremental rebrand/upgrade of Moonshot's existing WebBridge tool into a reusable-Skill browser extension, extending an already-known browser-agent pattern rather than shifting competitive dynamics.
Moonshot AI's Kimi Browser Extension Turns Web Actions Into Reusable Skills
Moonshot AI has turned its four-month-old Kimi WebBridge tool into a full Chrome/Edge browser extension, adding a sidebar chat interface and the ability to record repeated web actions as reusable "Skills." The extension still lets a Kimi agent open pages, read content, and click through a site inside the user's own browser, but now a workflow performed once, such as pulling the day's news from the same site and exporting it, can be saved and replayed with one click. The original mode, where a local agent drives the extension via API, remains available. Moonshot released the extension days after shipping Kimi Code Desktop, a project workspace with an embedded browser that lets a coding agent show its work and take point-and-click feedback on a page it just edited.
The move extends a browser-automation pattern that Manus (via Monica and Browser Operator) and ByteDance's Doubao plugin have already staked out, turning Kimi's agent-in-your-browser feature into something closer to a personal automation library. It lands while Moonshot is pushing distribution and revenue harder than product novelty: Kimi K3 went live on Amazon Bedrock this week, and the company is targeting $2B in annualized revenue by year-end ahead of a Hong Kong IPO reportedly seeking up to $50B in valuation. Per the AI Market Watch index, Moonshot-linked news volume in our pipeline rose to 45 items in the past 90 days from 28 in the prior period, coverage of pipeline-ingested sources only, consistent with a company in an active IPO and product push rather than a single standout release.
For builders, the reusable-Skill model is a low-cost way to make agentic browsing sticky for repeat workflows, RPA-style, without training a dedicated computer-use model. For investors watching the IPO story, splitting the agent surface into a project-bound coding environment and an everyday browser extension suggests Moonshot is building distribution breadth rather than a single flagship agent product, a hedge worth weighing against the revenue targets underpinning its valuation ask.
