chrome-mcp

name	chrome-mcp
description	Control Chrome browser efficiently with minimal token usage

Quick Usage

Navigate directly (don't use search boxes)

chrome_navigate_page({ url: "https://example.com/search?q=term" })

Extract data with evaluate_script (not screenshots)

chrome_evaluate_script({
  function: `() => {
    const items = document.querySelectorAll('a[href*="/item/"]');
    return Array.from(items).slice(0, 15).map(el => ({
      text: el.textContent?.substring(0, 150) || '',
      link: el.href
    }));
  }`
})

Take snapshot for page structure (lightweight)

chrome_take_snapshot()  // Returns a11y tree, good for finding elements

Take screenshot only for visual tasks

chrome_take_screenshot()  // Use sparingly - high token cost

---

Token-Efficient Patterns

1. URL Navigation > Form Filling

Good (1 call):

chrome_navigate_page({ url: "https://site.com/search?query=desk&maxPrice=300" })

Bad (3+ calls):

chrome_navigate_page({ url: "https://site.com" })
chrome_fill({ uid: "search-box", value: "desk" })
chrome_click({ uid: "search-button" })

2. evaluate_script > Snapshot for Data Extraction

Good - Get exactly what you need:

chrome_evaluate_script({
  function: `() => {
    return Array.from(document.querySelectorAll('.listing')).map(el => ({
      title: el.querySelector('h2')?.textContent,
      price: el.querySelector('.price')?.textContent,
      link: el.querySelector('a')?.href
    }));
  }`
})

Wasteful - Snapshot returns entire page structure:

chrome_take_snapshot()  // Then parse through huge response

3. Limit Results in Script

Always .slice() to avoid huge responses:

chrome_evaluate_script({
  function: `() => {
    return Array.from(document.querySelectorAll('a'))
      .slice(0, 20)  // Limit results
      .map(el => ({ text: el.textContent?.substring(0, 100), href: el.href }));
  }`
})

4. Truncate Text Content

// Good - truncated
text: el.textContent?.substring(0, 150) || ''

// Bad - could be huge
text: el.textContent

---

When to Use Each Tool

Tool	Use For	Token Cost
navigate_page	Go to URL, reload, back/forward	Low
evaluate_script	Extract data, click via JS, check state	Low
take_snapshot	Find element UIDs, understand page structure	Medium
take_screenshot	Visual comparison, verify appearance	High
click	Click element by UID (need snapshot first)	Low
fill	Fill form field (need snapshot first)	Low

Decision Tree

Need data from page?
├─ Structured data (links, prices, text)
│  └─ evaluate_script with targeted selectors
├─ Need to find element UIDs to click
│  └─ take_snapshot (once), then click/fill
└─ Visual comparison (style, color, condition)
   └─ take_screenshot

Need to navigate?
├─ Site has URL-based search/filters
│  └─ navigate_page with query params
└─ Must use forms
   └─ take_snapshot → fill → click

---

Common Selectors

Links with specific patterns

document.querySelectorAll('a[href*="/item/"]')      // Contains "/item/"
document.querySelectorAll('a[href^="/product"]')    // Starts with "/product"
document.querySelectorAll('a[href$=".html"]')       // Ends with ".html"

Extract price from mixed text

const priceMatch = text.match(/\$[\d,]+/);
const price = priceMatch ? priceMatch[0] : 'N/A';

Get page content for reading

chrome_evaluate_script({
  function: `() => ({ content: document.body.innerText.substring(0, 2000) })`
})

---

Multi-Account Sites (Gmail, Google)

Google uses /u/N/ for account indices:

https://mail.google.com/mail/u/0/  → First account
https://mail.google.com/mail/u/1/  → Second account
https://drive.google.com/drive/u/1/  → Second account's Drive

Navigate directly to correct account:

chrome_navigate_page({ url: "https://mail.google.com/mail/u/1/#inbox" })

---

Error Handling

Error	Solution
Navigation timeout	Usually OK - page loaded, continue
Element not found	Take fresh snapshot, UID may have changed
Empty results	Try broader selector or wait for page load
Login required	Tell user to log in manually in Chrome

Timeouts are Usually OK

// Even if this "times out", the page usually loaded
chrome_navigate_page({ url: "https://example.com" })
// Just continue with your next action

---

Page Management

List open pages

chrome_list_pages()

Switch to page by index

chrome_select_page({ pageIdx: 1 })

Open new page

chrome_new_page({ url: "https://example.com" })

Close page

chrome_close_page({ pageIdx: 0 })

---

Real Examples

Facebook Marketplace Search

// 1. Navigate directly with filters
chrome_navigate_page({ 
  url: "https://www.facebook.com/marketplace/sanfrancisco/search?query=standing%20desk&maxPrice=300" 
})

// 2. Extract listings efficiently
chrome_evaluate_script({
  function: `() => {
    const listings = document.querySelectorAll('a[href*="/marketplace/item/"]');
    return Array.from(listings).slice(0, 15).map(el => ({
      text: el.textContent?.substring(0, 150) || '',
      price: (el.textContent?.match(/\\$[\\d,]+/) || ['N/A'])[0],
      link: el.href
    }));
  }`
})

Gmail - Read Recent Emails

// 1. Navigate to inbox
chrome_navigate_page({ url: "https://mail.google.com/mail/u/0/#inbox" })

// 2. Get email subjects
chrome_evaluate_script({
  function: `() => {
    const rows = document.querySelectorAll('tr.zA');
    return Array.from(rows).slice(0, 10).map(row => ({
      from: row.querySelector('.yX')?.textContent || '',
      subject: row.querySelector('.y6')?.textContent || '',
      snippet: row.querySelector('.y2')?.textContent || ''
    }));
  }`
})

Check if Logged In

chrome_evaluate_script({
  function: `() => ({
    url: window.location.href,
    hasLoginButton: !!document.querySelector('[data-testid="login"]'),
    hasAvatar: !!document.querySelector('img[alt*="profile"], img[alt*="avatar"]')
  })`
})