Skip to main content
Yutori builds computer-use agents — agents that can reliably and autonomously execute tasks on a computer. We offer four APIs — Navigator, Browsing, Research, and Scouting. The Yutori Python SDK and MCP servers are also available for calling all Yutori APIs. The Playground runs the Navigator models against a hosted browser or desktop, so you can try them without writing code. Yutori Local is a Mac desktop app that gives our agents access to a browser on your computer, enabling login-required workflows and local browsing tasks. It drives a browser, so it runs the browser-use models. You can use platform.yutori.com to monitor your API usage, view task history, and manage billing. Technical blogs about our model and multi-agent orchestration are available here.
Navigator is our family of computer-use models. Each model takes the user’s instruction, the current screenshot, and the history of previous actions and screenshots, and predicts the next action — click, type, scroll, etc. You execute the returned action and send back the next screenshot. The API follows OpenAI’s Chat Completions interface. Navigator n2 is optimized for computer use — operating a full desktop, with shell access. Navigator n1.5 is optimized for browser use — interacting with webpages in browser environments. See the Navigator API to get started. If you don’t want to manage your own browser infra, see the Browsing API that runs Navigator models on our cloud browser.

​
Browsing API

The Browsing API enables automation of browser-based workflows. Simply describe your browser-based task in natural language, for example:
  • Fill a form on website
  • Check a webpage for changes and retrieve updated information
  • Check multiple webpages and download a summary in a structured format
  • Log into a website and enter or update information
and a Navigator-powered computer-use agent that runs its own cloud browser and operates it like a person will click, type, scroll, and navigate for you. For login/signup or other auth-heavy flows, we recommend setting "require_auth": true on your request so the system can prefer an auth-optimized browser provider. See the Browsing API to create your first browsing task.

​
Research API

The Research API enables one-time (wide and deep) research of anything on the web. Simply specify what you want to research in natural language, for example:
  • What are the latest developments in quantum computing?
  • Summarize recent AI model launches and their capabilities
  • Research competitor pricing for cloud GPU instances
Unlike the Browsing API (which uses a single navigator), Research API uses 100+ MCP tools for comprehensive web-based research. This provides the same research capabilities as the Scouting API, but without scheduling recurring runs.
See the Research API to create your first research task.

​
Scouting API

The Scouting API enables continuous monitoring of the web at a configurable schedule for tracking any changes relevant to a query. Simply specify what you’re looking to track in natural language, for example:
  • anytime a startup in SF announces seed funding
  • when H100 pricing per hour drops below $1.50
  • when a new AI model launches
and a ‘Scout’ will spin up a team of sub-agents to monitor — either specific URLs or the entire web — and alert you with structured data whenever there’s a relevant update.
While the Browsing and Research APIs are designed for one-time task execution — either by a single navigator in a cloud browser or by a multi-agent system — the Scouting API is meant for setting up periodic monitoring and alerts for changes over time.
See the Scouting API to create your first scouting task.
The Scouting API powers Yutori’s Scouts product. See how it works: