Vision Browser Agent
AI & KnowledgeAdvanced browser automation with vision, OCR, and optional multi-tab support
Overview
The Vision Browser Agent extends the basic browser loop with screenshot vision analysis, OCR, optional multi-tab browsing, and tenant-scoped persistent sessions. Use it when page content is visual, canvas-heavy, or spread across tabs.
Configuration
| Property | Type | Default | Description |
|---|---|---|---|
goalrequiredvars | multiline | — | Browsing task for the agent. |
startUrlvars | string | — | Optional starting URL. |
model | options | — | Primary reasoning model. |
visionModel | options | — | Model used for screenshot_vision analysis. |
vision | boolean | true | Enable vision and OCR actions. |
multiTab | boolean | false | Allow opening additional tabs (max 5). |
maxIterations | number | 15 | Maximum browser actions. |
timeout | number | 60000 | Per-action timeout in milliseconds. |
headless | boolean | true | Run Chromium headlessly. |
Output Variables
Reference these variables in downstream nodes using the {{node.field}} syntax.
{{node-id.result}}stringFinal answer
{{node-id.screenshots}}arrayCaptured screenshots
{{node-id.iterations}}numberSteps taken
{{node-id.session_id}}stringScoped session identifier when persistence is enabled
Examples
Visual pricing table
Use screenshot vision on a canvas-heavy pricing page.
{
"goal": "Extract the monthly and annual prices shown on the page",
"startUrl": "https://example.com/pricing",
"vision": true,
"multiTab": false
}Tips & Best Practices
- •Computer Use mode is not available yet and is blocked at runtime.
- •Custom JavaScript execution is disabled for security.
- •Enable multi-tab only when the task genuinely needs more than one tab.
Related Nodes
Try Vision Browser Agent in Your Workflow
Create a free account and start building AI-powered workflows with the Vision Browser Agent node.
Open Workflow Editor