Quickstart Guide

This guide will take you from zero to monitoring your first AI agent session in under 5 minutes.

Prerequisites

Chrome Browser: Version 134 or higher
AI Browser Agent: Claude, or other agent that operates in browser tabs

ContextFort automatically detects when agents like Claude enter “agent mode” and begin operating in your browser.

Step 1: Install ContextFort

Open Chrome Web Store

Navigate to the ContextFort extension page in the Chrome Web Store.

Add to Chrome

Click “Add to Chrome” button.Chrome will prompt you to confirm permissions:

Storage (for local screenshot and session data)
Tabs (to detect agent activity)
Web navigation (to track agent navigation)

Confirm Installation

Click “Add extension” to confirm. The ContextFort icon will appear in your Chrome toolbar.

Pin the extension: Click the puzzle piece icon in Chrome and pin ContextFort for easy access.

Step 2: Start an Agent Session

Now let’s trigger agent monitoring:

Open Claude or Your Agent

Navigate to your AI agent (e.g., claude.ai). Start a conversation.

Enable Agent Mode

Ask the agent to perform a browser task:

"Search Google for the latest news on AI safety"

When Claude enters agent mode, it will open a new tab with a tab group. The tab group title will show a ⌛ (hourglass) emoji.

ContextFort Activates

ContextFort automatically detects the agent tab group and begins monitoring:

Screenshots captured on every click and input
URLs visited are tracked
Session data is stored locally

The background worker (chrome-extension/background.js:730-749) handles agent detection and session creation.

First-Time Login Required: If the agent tries to access a domain where you’re logged in (like Gmail), ContextFort will prompt you to separate your human and agent sessions. See Session Isolation for details.

Step 3: View Your First Session

Open Dashboard

Click the ContextFort icon in your Chrome toolbar. This opens the main dashboard.Alternatively, you can navigate directly to:

chrome-extension://[extension-id]/dashboard/visibility/index.html

Navigate to Visibility

The dashboard opens to the Visibility tab by default. Here you’ll see:

Real-time screenshot feed
Action timeline (clicks, inputs, navigations)
URLs visited
Session metadata (start time, duration)

Explore Session Data

Click on any screenshot to see details:

Action Type: Click, input, page_read
Element Details: Tag name, class, ID
Timestamp: When the action occurred
URL: Page where action happened

Example event data structure (chrome-extension/background.js:1029-1052):

{
  id: 1740789123.456,
  sessionId: 1740789000000,
  tabId: 123,
  url: "https://google.com/search?q=AI+safety",
  title: "AI safety - Google Search",
  reason: "agent_event",
  timestamp: "2026-02-28T10:30:00.000Z",
  eventType: "click",
  eventDetails: {
    element: { tag: "BUTTON", className: "search-btn" },
    coordinates: { x: 450, y: 300 },
    actionType: "click"
  }
}

Step 4: Stop Agent Session

When the agent completes its task:

Agent Stops Naturally

The agent will typically stop on its own. Claude changes the tab group emoji from ⌛ to ✅.

ContextFort Ends Session

ContextFort detects the emoji change (chrome-extension/background.js:632-640) and:

Swaps agent cookies back to your human session
Ends the monitoring session
Calculates session duration
Keeps all screenshots and data in storage

View Session History

Navigate to Sessions tab in the dashboard to see:

All completed sessions
Session duration
Screenshot counts
Status (active/ended)

What Just Happened?

Here’s what ContextFort did behind the scenes:

Agent Detection (background.js:730-749)
- Content script detected agent activity via hourglass emoji
- Created new session with unique ID
- Started tracking tab group
Screenshot Capture (background.js:1059-1083)
- Captured visible tab on every click event
- Stored screenshot as data URL in Chrome storage
- Linked screenshot to session ID
Session Management (background.js:534-573)
- Tracked visited URLs for context mixing rules
- Maintained session state (active/ended)
- Calculated duration on session end
Cookie Isolation (background.js:156-202)
- Swapped your human cookies for agent cookies
- Prevented agent from accessing your logged-in accounts
- Restored human cookies when session ended

Next Steps

Configure Security Controls

Set up rules to block agents from sensitive domains

Understand Session Isolation

Learn how ContextFort separates human and agent sessions

Export Session Data

Download screenshots and session logs for review

Advanced Configuration

Customize ContextFort behavior and privacy settings

Troubleshooting

ContextFort didn't detect my agent

Make sure:

Agent opened a tab group with ⌛ emoji
Extension has necessary permissions
Chrome version is 134+
Check Chrome DevTools console for errors

No screenshots are appearing

Possible causes:

Agent didn’t perform any actions (only page loads)
Screenshot capture failed (check permissions)
Storage quota exceeded (clear old sessions)

Check storage usage:

chrome.storage.local.getBytesInUse(null, (bytes) => {
  console.log('Storage used:', bytes, 'bytes');
});

Agent was blocked unexpectedly

ContextFort may have enforced a security rule:

Check Controls > Blocking Rules for active rules
Look for context mixing violations (agent visited incompatible domains)
Review in-page notifications for block reasons

Blocking logic is in chrome-extension/background.js:1180-1239.

This happens when the agent needs a separate login profile:

Click “OK, let me login” in the notification
Log in with agent-specific credentials
Click “I’m logged in” to continue

ContextFort will remember this profile for future sessions. See chrome-extension/background.js:896-906 for the login flow.

Quick Reference

Extension Permissions

// chrome-extension/manifest.json:8-19
"permissions": [
  "storage",           // Store screenshots and sessions locally
  "unlimitedStorage",  // No storage limits for screenshots
  "tabs",              // Detect agent tabs
  "tabGroups",         // Monitor tab group emojis
  "webNavigation",     // Track agent navigations
  "declarativeNetRequest", // Enforce blocking rules
  "scripting",         // Inject content scripts
  "cookies",           // Session isolation (swap cookies)
  "contextMenus",      // Right-click menu options
  "notifications"      // Login notifications
]

Keyboard Shortcuts

Shortcut	Action
`Ctrl+Shift+S` (Windows)	Snapshot current session
`Cmd+Shift+S` (Mac)	Snapshot current session

Keyboard shortcut defined in chrome-extension/manifest.json:53-60.

Getting Help

If you’re stuck:

Check the FAQ for common issues
Review GitHub Issues for known bugs
Open a new issue with:
- Chrome version
- Agent type (Claude, etc.)
- Steps to reproduce
- Console errors (if any)

Ready to explore?

Learn about all ContextFort features and capabilities.

Get Started

Core Features

Dashboard

Configuration

Security & Compliance

Quickstart

Quickstart Guide

Prerequisites

Step 1: Install ContextFort

Step 2: Start an Agent Session

Step 3: View Your First Session

Step 4: Stop Agent Session

What Just Happened?

Next Steps

Configure Security Controls

Understand Session Isolation

Export Session Data

Advanced Configuration

Troubleshooting

Quick Reference

Extension Permissions

Keyboard Shortcuts

Getting Help

Ready to explore?

Build docs developers (and LLMs) love

Get Started

Core Features

Dashboard

Configuration

Security & Compliance

​Quickstart Guide

​Prerequisites

​Step 1: Install ContextFort

​Step 2: Start an Agent Session

​Step 3: View Your First Session

​Step 4: Stop Agent Session

​What Just Happened?

​Next Steps

Configure Security Controls

Understand Session Isolation

Export Session Data

Advanced Configuration

​Troubleshooting

​Quick Reference

​Extension Permissions

​Keyboard Shortcuts

​Getting Help

Ready to explore?

Build docs developers (and LLMs) love

Quickstart Guide

Prerequisites

Step 1: Install ContextFort

Step 2: Start an Agent Session

Step 3: View Your First Session

Step 4: Stop Agent Session

What Just Happened?

Next Steps

Troubleshooting

Quick Reference

Extension Permissions

Keyboard Shortcuts

Getting Help