Skip to main content

Vercept

Category: AI Agents

Seattle-based AI startup building computer vision-powered desktop agents that see and act on computers like humans do, acquired by Anthropic in 2026. Vercept was founded in 2024. The company is led by Kiana Ehsani. Based in Seattle, Washington, USA. Team size: 1-10. Total funding raised: $50.0M. Latest round: Seed. Key investors include Fifty Years (lead), Point Nine Capital, AI2 Incubator, Madrona Venture Group, Eric Schmidt (angel), Jeff Dean (angel), Kyle Vogt (angel), Arash Ferdowsi (angel).

Founded
2024
Headquarters
Seattle, Washington, USA
Team size
1-10
Total funding
$50.0M

Value proposition

AI that sees and acts on computers like humans do — no APIs, no browser extensions, no hardcoded integrations. A vision-first approach to desktop automation that works across any application.

Products and solutions

Vy (Mac desktop AI agent that sees and interacts with computer screens via natural language), VyUI (proprietary computer vision model for UI grounding, achieving SOTA on ScreenSpot benchmarks), API for developers to build computer-use agents, UI test suites, and RPA solutions.

Unique value

Vision-first desktop AI agent that understands screens visually (not via APIs/HTML), achieving 92% accuracy on ScreenSpot v1 vs OpenAI's 18.3%, and operating across any application on Mac without pre-built connectors.

Target customer

Enterprise customers in finance, software development, and content creation; individual power users; accessibility users (speech-to-text integration); businesses automating multi-step workflows across applications.

Industries served

Finance, Software Development, Content Creation, Accessibility, Enterprise Automation

Technology advantage

VyUI — proprietary computer vision model achieving state-of-the-art on UI grounding benchmarks (ScreenSpot v1: 92%, ScreenSpot v2: 94.7%, GroundUI Web: 84.8%); vision-first architecture requiring no APIs or browser extensions; built by world-class AI researchers from AI2 with deep expertise in computer vision, agents, and reinforcement learning.

How they differentiate

Unlike traditional RPA (UiPath, Automation Anywhere) that requires hardcoded interactions and pre-built connectors, Vy uses a vision-first approach to understand any screen visually. Outperformed OpenAI, Claude, and Gemini on computer-use benchmarks. Built by elite AI researchers from the Allen Institute for AI with deep computer vision expertise.

Main competitors

UiPath, OpenAI Operator, Google Project Mariner, Simular, Anthropic Claude Computer Use (post-acquisition)

Key partnerships

Together AI (inference infrastructure partner — deployed VyUI models on Together's platform for 5x performance improvement), AI2 Incubator (first institutional investor and incubator), Anthropic (acquirer, 2026)

Notable customers

Enterprise customers across finance, software development, and content creation industries (specific names not publicly disclosed)

Major milestones

Founded 2024 by former AI2 researchers, Emerged from stealth Feb 2025, Raised $16M seed round Jan 2025 at $67M post-money valuation, Launched Vy Mac app April 2025, VyUI achieved SOTA on ScreenSpot benchmarks, Raised total of $50M+, Acquired by Anthropic Feb 2026 to advance Claude's computer-use capabilities

Growth metrics

VyUI achieved 92% accuracy on ScreenSpot v1 vs OpenAI's 18.3%; 94.7% on ScreenSpot v2; 84.8% on GroundUI Web; 9-person team at time of acquisition

Market positioning

Early leader in vision-based desktop AI agents, competing with both traditional RPA platforms (UiPath) and AI-native computer-use tools (OpenAI Operator, Google Project Mariner, Amazon Nova Act, Simular). Acquired by Anthropic to advance Claude's computer-use capabilities.

Geographic focus

Global (headquartered in Seattle, USA)

Patents and IP

VyUI — proprietary computer vision model for UI grounding and screen understanding (specific patent details not publicly disclosed)

About Kiana Ehsani

Ex-Senior Researcher at Allen Institute for AI (AI2), leading robotics and embodied AI teams; PhD in Computer Science from University of Washington; internships at Google and Meta.

Official website: