Yutori Browser Use

Build browser-use agents with Yutori n1 — state-of-the-art browser-use model

n1 is a browser-use model developed by Yutori, which enables building computer use agents (CUA). Given a user goal, a webpage screenshot, and action history, n1 predicts browser actions — such as clicking, typing, scrolling — to complete complex multi-step tasks. The model is trained on large-scale real-world web interaction trajectories and reinforcement learning with simulated and real websites.

In Anchor Browser, this agent is available directly through agentic API’s and taskOptions.agent = 'yutori'.

Overview

Yutori Browser Use provides:

• Browser-use automation with native browser actions
• State-of-the-art performance — outperforms GPT 5.4 and Opus 4.6 (+3%) on Navi-bench and WestWorld, while being 2.5x faster and 4-5x cheaper
• Multi-step task handling for navigation, extraction, and workflow completion
• Structured output support through output_schema when you need machine-readable results

Learn more about the model in the official Yutori n1 reference and the Introducing n1 blog post.

Supported Models

Model	Model ID	Best For
Yutori n1	n1-latest	State-of-the-art browser-use automation (default)

Code Example

node.js

import Anchorbrowser from 'anchorbrowser';

const anchorClient = new Anchorbrowser({
  apiKey: process.env.ANCHORBROWSER_API_KEY,
});

const response = await anchorClient.agent.task(
  'Go to Hacker News, open the top story, and summarize it in 3 bullet points',
  {
    taskOptions: {
      url: 'https://news.ycombinator.com',
      agent: 'yutori',
    },
  }
);

console.log(response);

Supported Browser-Use Actions

Yutori in Anchor Browser supports these core browser-use actions:

• left_click, double_click, triple_click, right_click
• scroll, scroll_at
• type, key_press
• hover, drag
• goto_url, go_back, refresh
• wait

Best Practices

• Set max_steps intentionally based on task complexity to control runtime
• Use output_schema for reliable, structured extraction workflows
• Write specific prompts with clear success criteria for better automation quality

Secure Credentials with Secret Values

When using Yutori n1 on Anchor Browser, you get secure credential handling built into the infrastructure. Secrets are never exposed to the AI model.

Learn more about domain-scoped secrets and TOTP support.