Browser Memory provides reusable browser Skills for AI agents, eliminating the need to rediscover websites on every session. Instead of exploring the DOM from scratch, agents can download and execute community-created Skills that encapsulate navigation, selectors, and workflows. Reduce exploration time and token usage in Claude Code, Codex, Cursor, OpenHands, and any browser agent.

I've been using browser agents for a while. The problem I kept running into is that the agent has to re-learn a website every time it visits it. That makes browsing slow and wastes a lot of tokens. So I'm building Browser Memory: an open, shared catalog of browser Skills for every website on the web. Instead of figuring out a site from scratch, your agent simply downloads a ready-to-run Skill from the catalog—log in, search, extract data, and more. When an agent solves a new workflow, it can publish that Skill back to the catalog, so everyone benefits and the ecosystem keeps growing. Open source and works with Claude Code, Codex, Cursor, OpenHands, and any browser agent. GitHub: https://github.com/browser-memor... Site: https://browser-memory.com/

A versioned trust system would be great, like showing how often a Skill has been downloaded, verified by maintainers, or flagged broken so I know which ones to grab first. A simple badge or score would help me pick reliable ones without testing every one myself.

browser-memory

Community browser skills for AI agents. Save time and token

Community browser skills for AI agents. Save time and token