I built a browser for AI agents, then benchmarked it against Playwright MCP and browser-use

说明与用途

AgentGo 浏览器使用紧凑的页面概述和差异更新，减少了令牌消耗和执行时间，在 MiniWoB++ 任务中优于 browser-use 和 Playwright MCP。

问题

AI agents reading too much of a page snapshot, wasting tokens on unchanged content

结果

AgentGo solved 22/24 MiniWoB++ tasks, median 32 s, 94.7k input tokens, 7.7 turns, outperforming browser-use (19/24, 92 s, 209.9k tokens, 13.2 turns) and Playwright MCP (18/24, 42 s, 178.4k tokens, 13.0 turns)

工作流

• 1. Agent opens page with browser_open, receiving a compact outline of headings and actionable elements
• 2. Agent performs actions via browser_act
• 3. Browser returns only the diff (+, -, ~) of changed elements
• 4. Browser waits for page to settle before responding, eliminating unnecessary wait actions

证据

Every agent I tried on the web had the same problem: it was reading too much. A browser tool hands the model a page snapshot, the model clicks, and then gets a fresh snapshot of the whole page again.

来源

Same model (gpt-6-luna, medium reasoning), same 24 BrowserGym MiniWoB++ tasks, a fresh agent session for every task, only that browser's MCP tools, and the page's own reward decides.

来源

AgentGo 0.1.5 22/24 32 s 94.7k 7.7

来源

Three design choices do most of the work: 1) A compact outline instead of a snapshot. 2) Only what changed. 3) The browser waits, not the model.

来源