Config is the same across clients — only the file and path differ.
{
"mcpServers": {
"io-github-matthewalexong-neo-vision": {
"args": [
"-y",
"neo-vision"
],
"command": "npx"
}
}
}Are you the author?
Add this badge to your README to show your security score and help users find safe servers.
Give your AI agent a pixel-precise JSON map of every element on a page — coordinates, ARIA roles, accessible labels, and actionability flags — without screenshots, without brittle CSS selectors, without getting blocked by anti-bot systems.
Run this in your terminal to verify the server starts. Then let us know if it worked — your result helps other developers.
npx -y 'neo-vision' 2>&1 | head -1 && echo "✓ Server started successfully"
After testing, let us know if it worked:
Five weighted categories — click any category to see the underlying evidence.
No known CVEs.
Checked neo-vision against OSV.dev.
Click any tool to inspect its schema.
Be the first to review
Have you used this server?
Share your experience — it helps other developers decide.
Sign in to write a review.
Others in browser
Browser automation with Puppeteer for web scraping and testing
🔥 Official Firecrawl MCP Server - Adds powerful web scraping and search to Cursor, Claude and any other LLM clients.
MCP server for Firecrawl — search, scrape, and interact with the web. Supports both cloud and self-hosted instances. Features include web search, scraping, page interaction, batch processing, and LLM-powered content analysis.
The Apify MCP server enables your AI agents to extract data from social media, search engines, maps, e-commerce sites, or any other website using thousands of ready-made scrapers, crawlers, and automation tools available on the Apify Store.
MCP Security Weekly
Get CVE alerts and security updates for io.github.matthewalexong/neo-vision and similar servers.
Start a conversation
Ask a question, share a tip, or report an issue.
Sign in to join the discussion.
See the web the way Neo sees the Matrix.
Give your AI agent a pixel-precise JSON map of every element on a page — coordinates, ARIA roles, accessible labels, and actionability flags — without screenshots, without brittle CSS selectors, without getting blocked by anti-bot systems.
Version 0.6.0 · MIT License · GitHub
AI agents navigating the web today are stuck between two bad options:
Meanwhile, anti-bot systems block headless browsers on sight. So even if you solve the navigation problem, you can't get past the front door of sites like Yelp, LinkedIn, or Zillow.
NeoVision asks the browser's own layout engine where everything is — because it already knows. Like Neo seeing through the green code to perceive the real world, NeoVision reads the raw DOM but gives your agent a spatial map with ground-truth pixel coordinates, straight from the rendering engine.
{
"tag": "button",
"role": "button",
"label": "Sign in",
"bounds": { "x": 305, "y": 510, "width": 74, "height": 36 },
"click_center": { "x": 342, "y": 528 },
"actionable": true
}
No guessing. No hallucination. No selector that breaks tomorrow.
NeoVision drives the user's real Chrome via a Chrome extension — with real cookies, real fingerprint, real browsing history. Anti-bot systems see a real user because it is a real browser.
We tested this against the five most notoriously anti-bot sites on the web — Ticketmaster, Nike, LinkedIn, Instagram, and Amazon — plus Discord (Cloudflare). All six returned full page content with zero CAPTCHAs, zero bot walls, and zero detection signals. Full test report →
chrome://extensionsextension/ folder from this repobrew install cliclick
spatial_click and spatial_type dispatch real CGEvents through cliclick by default. Without it, those two tools will return a helpful error pointing you here. The other tools (snapshot, query, scroll, screenshot, etc.) work without cliclick.
{
"mcpServers": {
"neo-vision": {
"command": "npx",
"args": ["neo-vision"]
}
}
}
The MCP server starts automatically. The Chrome extension connects to it via WebSocket. Once connected, all 16 spatial tools are available.
NeoVision exposes 16 tools through the MCP server, organized by function:
| Tool | Description |
|---|---|
spatial_snapshot | Navigate to a URL and return a spatial DOM map with element coordinates, ARIA roles, and actionability flags. Supports compact and agent output formats. Now CSP-safe — works on x.com, GitHub, banks, and other strict-CSP sites. |
spatial_click | Real OS-level mouse click via cliclick (CGEvent, event.isTrusted = true). Cursor visibly travels to the target with eased animation; ±3px coord jitter; post-arrival pause. Defeats isTrusted-based anti-bot detection. Pass synthetic: true for the legacy in-page MouseEvent dispatch (only for edge cases). Requires brew install cliclick. |
spatial_type | Real OS-level keystrokes via cliclick. Per-keystroke 60–180ms delays + occasional thinking pauses on word boundaries. event.isTrusted = true. Auto-focuses the field by clicking it first if x/y given. Requires brew install cliclick. |
spatial_scroll | Scroll the page or a specific scrol |