Browser-driving agent that completes web tasks autonomously — booking, shopping, research.
BrowserTool useVisionMemory
Demo · hover to play
18 agents · capability hub
AI agents with vision2026
Agents that see — read screenshots, parse charts, understand UI layouts, interpret diagrams. Required for any agent that interacts with software made for humans.
Want the technical definition? Read the vision glossary entry →
The 18 agents that ship vision
Microsoft's AI work assistant — agents across Word, Excel, Outlook, Teams, and the Microsoft 365 stack.
Tool useRAGMemoryVisionDemo · hover to playClaude with computer-use capability — sees the screen, moves cursor, types, navigates apps autonomously.
BrowserTool useVisionMemory
Demo · hover to playVercel's generative UI agent — design and ship React components from natural language.
CodeTool useVisionDemo · hover to playAI video studio — turns scripts into polished talking-head videos with avatars in 140+ languages.
VisionVoiceTool useDemo · hover to playAI video avatars — turn text or audio into talking-head clips with photorealistic presenters.
VisionVoiceTool useDemo · hover to playAI video generation studio for creators — text-to-video, image-to-video, and full directorial control.
VisionTool useDemo · hover to playVibe-coding builder for non-engineers — prompt a full-stack app and ship it to a live URL in minutes.
CodeTool useVisionDemo · hover to playStackBlitz's in-browser AI builder — generates and deploys real Node.js apps from a single prompt.
CodeTool useVisionDemo · hover to playAI video avatar agent — turns a script into a studio-quality talking-head video in any language.
VisionVoiceTool useDemo · hover to playLong-running researcher inside Gemini that plans, browses and writes briefs.
BrowserRAGMemoryVision
Demo · hover to playDigital humans for customer interactions — autonomous animated characters with realistic emotion.
VoiceVisionMemoryDemo · hover to playPersonal AI agent that browses the web for you — books flights, fills forms, completes tasks autonomously.
BrowserTool useVisionMemoryDemo · hover to playPersonalized AI video at scale — clones one video into thousands tailored to each viewer.
VisionVoiceTool useDemo · hover to playAutonomous AP accountant — reads invoices, codes GL accounts, routes approvals, posts to your ERP.
VisionTool useMemoryDemo · hover to playAd-creative agents that generate and AB-test full video campaigns.
VisionTool useMemoryBrowser-driving AI agent — completes multi-step workflows on real web apps the way a human would.
BrowserTool useMemoryVisionVisual canvas agent — plans, drafts and thinks alongside you.
Tool useMemoryVisionDemo · hover to play
Frequently asked
What is vision in AI agents?+
An agent capability for understanding images, screenshots, and video — letting the model reason over visual content.
Which AI agents support vision?+
18 agents in our index ship vision. The list above is sorted by community interest; OpenAI Operator, Microsoft Copilot, Anthropic Computer Use are the most-researched in 2026.
How do I evaluate vision in an AI agent?+
Look for: (1) reliability across edge cases, not just demo videos; (2) how the agent recovers when vision fails mid-task; (3) whether vision is the default mode or an opt-in feature. Production-ready agents publish vision benchmarks; demos and screenshots aren't enough.
Explore other capabilities