Releases·Tool Release
Hyper-Browsing: open-source Playwright agent skill reuses learned site workflows to cut DOM exploration and tokens
A developer released Hyper-Browsing, an open-source agent skill built on Playwright and the user's local Chrome session. It lets agents inspect pages with built-in tools, then save successful workflows as reusable site knowledge so repeat visits skip DOM exploration. The author reports roughly 10x faster task completion on repeated tasks, though this is a subjective observation, not a benchmark.
2 engagement1 sourceTue, Sep 22, 2026, 12:10 AM
AI Overview
- Hyper-Browsing is an open-source agent skill built on Playwright that uses a local Chrome session instead of a hosted browser
- On new sites, built-in tools inspect forms, handle dynamic pages, scroll virtual lists, detect state changes, and extract only needed DOM parts
- Successful workflows are saved as site knowledge plus reusable scripts, letting agents skip DOM exploration on repeat visits
- Compact DOM inspection reduces agent context and token usage, with fallback to discovery when a saved workflow breaks after site updates
- The author reports roughly 10x faster repeated tasks than driving plain Playwright from scratch, flagged as subjective, not a benchmark
Entities
Related
Social·Community
Reddit user seeks fix for browser agents repeating navigation failures across sessions
A developer using the Hermes agent with a Camoufox browser backend reports that agents repeatedly fail on the same UI interactions, such as opening web app menus, and re-attempt the same JS-based fixes each run. The post asks how to make agents iteratively learn from completed steps without losing generality, noting the failures waste tokens and slow tasks.
A developer using the Hermes agent with a Camoufox browser backend reports that agents repeatedly fail on the same UI interactions, such as opening web app menus, and re-attempt the same JS-based fixes each run. The post asks how to make agents iteratively learn from completed steps without losing generality, noting the failures waste tokens and slow tasks.
8 engagement1 sourceredditSat, Oct 10, 2026, 11:52 AM
Products·launched
PawBrowse - faster agent browser automation via element tables instead of screenshots
PawBrowse is an open source browser automation driver for LLM agents. Instead of screenshot-based perceive-then-act loops, it gives the agent a compact table of clickable elements (accessible names, values, state) refreshed after each action, halving round-trips. It targets Claude Code users building browser-driving agents.
PawBrowse is an open source browser automation driver for LLM agents. Instead of screenshot-based perceive-then-act loops, it gives the agent a compact table of clickable elements (accessible names, values, state) refreshed after each action, halving round-trips. It targets Claude Code users building browser-driving agents.
17 engagement1 sourceredditSun, Sep 27, 2026, 11:47 AM
Releases·Agent Release
Jev Ultrafast: A Browser Agent That Can Navigate Websites in Seconds
Jev Ultrafast: A Browser Agent That Can Navigate Websites in Seconds
Jev Ultrafast: A Browser Agent That Can Navigate Websites in Seconds
0 engagement1 sourcerssFri, Sep 25, 2026, 04:38 AM
Products·launched
Public Browser – MCP server running real Chrome with your profile via CDP for AI agents
Public Browser is an open-source MCP server that lets LLM agents control a real Chrome browser using the user's existing profile over the CDP protocol, with no separate logins. It targets developers frustrated with token waste and slowness of Playwright or Chrome DevTools-based agent browsing. The author reports benchmark gains versus AgentBrowser in calls, tokens, cost, and time.
Public Browser is an open-source MCP server that lets LLM agents control a real Chrome browser using the user's existing profile over the CDP protocol, with no separate logins. It targets developers frustrated with token waste and slowness of Playwright or Chrome DevTools-based agent browsing. The author reports benchmark gains versus AgentBrowser in calls, tokens, cost, and time.
1 engagement1 sourcehackernewsThu, Sep 24, 2026, 09:20 PM
Products·launched
Pickle: Token-efficient AI agent browser with policy-gated actions
Pickle is a browser designed for AI agents that loads webpages as compact structured data instead of raw HTML, reducing token usage. It includes standard browser features like tabs, search history, and bookmarks, and adds policy gating for actions like purchases, with all actions logged for transparency. It solves the problem of high token costs and lack of control when agents browse the web.
Pickle is a browser designed for AI agents that loads webpages as compact structured data instead of raw HTML, reducing token usage. It includes standard browser features like tabs, search history, and bookmarks, and adds policy gating for actions like purchases, with all actions logged for transparency. It solves the problem of high token costs and lack of control when agents browse the web.
3 engagement1 sourcehackernewsTue, Aug 11, 2026, 03:43 PM