Tag
fastbrowse 是一款开源浏览器代理，能读取页面上的按钮、链接和输入框，并由 System One 模型选择下一步操作，不确定时回退到标准模型，主打快速、低成本且每条回答都附带引用。
开发者介绍 Talos/Taloswork 浏览器代理的设计思路：首次访问用轻量无障碍快照加 LLM 理解页面，后续复用站点结构配方，并在遇到 CAPTCHA 或结构变化时交由人工接管，同时征集直接的产品反馈。
Retriever AI 推出免费 AI 浏览器代理，可在 Chrome 或云端浏览器中自动完成找潜在客户、发送外联、筛选创作者及申请职位等跨网站任务，宣称在准确率上超越 OpenAI、Anthropic 等模型，速度快 5 倍、成本低 300 倍。
作者用单张 3090 上运行的本地 35B 浏览器 agent（Ornith 1.5）在 Online-Mind2Web 上跑了部分任务，通过三轮 harness/清单提示改进将一个错误任务修好，并报告了 30 个新任务的结果：研究类任务表现尚可，UI 操作（下拉框、cookie 遮罩、位置自动补全）是主要失分点。
A developer is building a browser-agent Chrome extension as an alternative to Claude in Chrome and is asking users for feedback on pain points and desired features.
The author built 'hunch', an open-source browser agent tool that uses a lightweight model (jev) for quick ~150ms decisions per step, reducing latency compared to LLM calls and integrating under existing agents like Claude Code with structured exit reports.
Ultrafast, an open-source browser agent, gained 15,000 GitHub stars in four days, highlighting its capabilities in fast and low-cost automation of tasks like flight searches using LLMs.
Browser Use and Jev developed an ultra-fast open-source browser agent that can perform tasks like searching for flights in 7 seconds at a cost of just $0.0039, making AI-driven browser automation more practical and efficient.
onPanda is an interactive tool for token-level visualization and control of LLMs and agents, featuring data annotation, model inspection, and support for multiple modalities including browser-based execution.
A tweet announces the release of an ultrafast, opensource browser agent, humorously comparing its availability to the anticipated launch of GTA 6.
The article discusses a common failure in browser agents where form rejections are not detected due to reliance on structural changes in action results, and suggests including visible text in observations and treating refusal as a first-class outcome to improve reliability.
Retriever AI has launched a totally free browser agent extension supported by ads, leveraging DeepSeek Flash Code Mode to minimize costs and enable continuous learning from user workflows.
The article discusses challenges with browser agents in data extraction, such as incomplete tables, and introduces Thunderbit's solution that provides source URLs and reviewable tables for transparency and trust.
A developer shares how they cut DeepSeek API costs by roughly 90% through prefix caching optimizations, reducing browser agent task costs below $0.005 and enabling a free, ads-supported browser agent.
WebRider is a hierarchical framework that formalizes delegated web tasks as intent contracts, preserving persona-conditioned policies through every browsing step. It includes RiderBench, a benchmark of 4,096 live-web contracts, and an 8B action-policy model trained through its guarded interface.
Hark previews Hark Handoff, a browser-use agent that can complete tasks like shopping and booking reservations by navigating websites without APIs, with claims of being faster and cheaper than rivals.
Index is an open-source AI browser Agent developed by the Laminar team that turns any website into a callable API, supports vision models such as Claude and Gemini, and achieves 92% accuracy on WebVoyager.
Microsoft released Fara1.5-27B, a 27B parameter multimodal computer use agent for web browsers, fine-tuned from Qwen3.5-27B, capable of end-to-end web tasks via screenshots.
barebrowse is a tool that generates pruned ARIA snapshots from web pages, reducing token consumption for local AI models by stripping boilerplate and ads, and reusing existing browser cookies.
A browser-based agent that reverse-engineers web app APIs to automatically generate agent tools, enabling deep integration without code changes.