Composite enters the AI agent sector with a distinct philosophy: the best browser is the one a user already uses. While competitors like OpenAI with its Atlas browser and The Browser Company with its Dia browser require users to migrate their entire digital workflow to a standalone application, Composite is a cross-browser extension for Chrome, Edge, and Brave. This architectural choice addresses a significant friction point in the agent market: the high cost and inconvenience of switching core environments.

The company was founded in San Francisco in 2024 by Yang Fan Yun, a former product manager at Uber, and Charlie Deane. In late 2025, they secured a $5.6 million seed round led by Menlo Ventures, with participation from Nat Friedman and Daniel Gross. This backing highlights Composite as a prominent player in the agentic wave of AI startups that prioritize task execution over conversational assistance.

Autopilot for the DOM

Composite is an AI worker that navigates the Document Object Model (DOM). While many AI browser extensions offer a sidebar for summarizing text or drafting emails, Composite is designed to take actions. Using a keyboard shortcut—Cmd + Shift + Space—users invoke a Spotlight view where they can describe a task in plain English. The agent then navigates sites, clicks buttons, extracts data, and fills forms. This capability is useful for professional workflows where official APIs do not exist. Instead of waiting for a native integration between two SaaS platforms, a user can instruct Composite to move data from a LinkedIn profile to an Applicant Tracking System or to triage tickets between Jira and Salesforce.

Multi-model routing and local execution

The platform uses a multi-model routing system to manage the latency and cost of large language models. Simple navigational steps are routed to faster, specialized models, while more complex reasoning or UI interpretation tasks are sent to capable vision-based models. This execution happens locally via a companion desktop application. By operating within the user's existing browser session, Composite leverages active logins and cookies. This avoids the security risk of sharing sensitive credentials or API keys with a third-party server, a common hurdle for enterprise adoption of AI agents.

The shift to professional operations

The competitive market for Composite is split between tech giants and smaller automation tools. On one side, Google’s Gemini and OpenAI’s Atlas focus on consumer use cases like travel booking. On the other side are browser-based automation tools like HARPA AI or Axiom. Composite positions itself between these poles as a tool for operations and recruiting teams. Their marketing targets enterprise users who require parallel execution—running multiple automation threads simultaneously—and strict compliance through a zero data retention policy.

While the product remains in beta, the focus is clearly on professional teams. By automating the repetitive tasks that populate the workday of an operations manager or a recruiter, Composite aims to move AI agents from the category of "shopping assistants" to essential business infrastructure.