mcpField
N
Neo Vision

Neo Vision

Local
github.com/matthewalexongBrowser Automation0
Repository

About Neo Vision

Give your AI agent a pixel-precise spatial map of every element on a web page — coordinates, ARIA roles, accessible labels, and actionability flags — without screenshots, without brittle CSS selectors, without getting blocked by anti-bot systems. Built-in stealth layer patches

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
  "mcpServers": {
    "neo-vision": {
      "command": "npx",
      "args": [
        "-y",
        "neo-vision"
      ]
    }
  }
}
Run commandnpx -y neo-vision

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
npmneo-vision0.2.2

Tools

5

Saved tool list · updated 3w ago

ParameterTypeRequiredDescription
urlstringYesURL to navigate to. A real Chrome browser is launched automatically — no setup needed.
viewport_widthintegerNoViewport width in CSS pixels (default: 1280)
viewport_heightintegerNoViewport height in CSS pixels (default: 720)
zoomnumberNoDevice scale factor / zoom level (default: 1.0)
settle_msintegerNoMilliseconds to wait after page load for dynamic content (SPAs, lazy-loading) to settle (default: 2000)
include_non_visiblebooleanNoInclude hidden elements (display:none, visibility:hidden). Usually not needed.
max_depthintegerNoMax DOM depth to traverse (default: 50). Increase only for deeply nested pages.
verbosityenumNoWhat to include: 'actionable' = buttons/links/inputs only (default, fastest), 'landmarks' = + headings/nav/sections, 'all' = every visible elementactionable | landmarks | all
max_elementsintegerNoMaximum elements to return (default: 2000). If truncated, scroll + re-snapshot to see more.
compactbooleanNoReturn compact elements (default: true). Compact strips CSS layout details, selectors, and DOM hierarchy — keeping only what you need: tag, role, label, text, bounds, click_center, actionable. Set false for full element data.
output_formatenumNoOutput format: 'compact' (default) returns the full element list with coordinates, roles, and accessibility info for general browsing and UI tasks. 'agent' is optimized for AI context windows on text-dense pages (Wikipedia, Amazon, long articles): returns deduplicated readable page text + only interactive elements within the current viewport with click coordinates. In agent mode, elements are strictly confined to viewport bounds using click_center checks (not just bounding box overlap), preventing wide-layout leakage. Includes viewport_info with scroll position, page height, and percent_remaining. For long pages: use spatial_scroll(delta_y=~670) to advance, then re-request agent output for the next viewport.compact | agent

Details

The maintainer has not added details yet. Check the README in the repository.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.