Skip to content
Gee-Code

Web & Browser

Gee-Code can access the web, search for information, and automate browsers — both headless browsers for automation and Chrome DevTools for debugging running apps.

Web Fetching

Section titled “Web Fetching”

WebFetch

Section titled “WebFetch”

Fetch and process content from any URL:

WebFetch(url="https://docs.python.org/3/library/asyncio.html", prompt="Summarize the key concepts")
  • Converts HTML to markdown for clean processing
  • Uses an AI model to extract the information you need
  • 15-minute cache for repeated access to the same URL
  • Follows redirects automatically

WebSearch

Section titled “WebSearch”

Search the web for information:

WebSearch(query="python async best practices 2026")

If WebSearch isn’t configured locally, use the Gee backend:

GeeConnect(service="search", action="web", params={"query": "python async patterns"})

Browser Automation (AgentBrowser)

Section titled “Browser Automation (AgentBrowser)”

The AgentBrowser is a headless Playwright-based browser for automating web interactions.

Workflow

Section titled “Workflow”
  1. Open a page
  2. Snapshot to get element references
  3. Interact using those references
AgentBrowser(command="open https://example.com")
AgentBrowser(command="snapshot")                    # Returns @e1, @e2, etc.
AgentBrowser(command="fill @e3 \"user@email.com\"")
AgentBrowser(command="click @e5")
AgentBrowser(command="screenshot /tmp/result.png")

Commands

Section titled “Commands”
CommandDescription
open <url>Navigate to a URL
snapshotGet accessibility tree with element refs
screenshotTake a screenshot
click <ref>Click an element
fill <ref> <text>Fill an input field
type <ref> <text>Type into an element (appends)
scroll up/downScroll the page
press <key>Press a keyboard key
waitWait for a condition
closeClose the browser

Use Cases

Section titled “Use Cases”
  • Filling out web forms
  • Automating repetitive web tasks
  • Testing web applications
  • Scraping dynamic content
  • Taking visual snapshots of pages

DevBrowser (Chrome DevTools)

Section titled “DevBrowser (Chrome DevTools)”

Connect to a running Chrome instance to debug web applications in real time.

Setup

Section titled “Setup”

Start Chrome with remote debugging enabled:

DevBrowserStartChrome(url="http://localhost:3000")
DevBrowserConnect()

This launches a separate Chrome profile (doesn’t affect your normal browser).

Debugging Tools

Section titled “Debugging Tools”
ToolDescription
DevBrowserConsoleRead console logs, warnings, and errors
DevBrowserDomInspect DOM structure by CSS selector
DevBrowserEvalExecute JavaScript in the page context
DevBrowserScreenshotCapture page or element screenshots
DevBrowserNetworkMonitor network requests and responses
DevBrowserClickClick elements by CSS selector
DevBrowserFillFill inputs (triggers React/Vue/Angular events)
DevBrowserReloadReload the page

Typical Workflow

Section titled “Typical Workflow”
  1. Start your dev server: npm run dev
  2. Launch debug Chrome: DevBrowserStartChrome(url="http://localhost:3000")
  3. Connect: DevBrowserConnect()
  4. Check console: DevBrowserConsole() — see errors and warnings
  5. Inspect DOM: DevBrowserDom(selector=".error-banner")
  6. Fix code: Edit the source files
  7. Reload: DevBrowserReload()
  8. Verify: DevBrowserScreenshot()

This lets the AI debug visual issues, runtime errors, and API failures by actually looking at your running app.

Next Steps

Section titled “Next Steps”