Gen AI Browser Integration for Prompt Security (Edge)

The AccuKnox Prompt Firewall browser plugin intercepts prompts and responses directly in ChatGPT, Claude, GitHub Copilot, and Gemini before they leave your browser session. This includes blocking prompt injection attacks, where a malicious input tries to override the model's behavior. Blocked prompts show a red banner. Allowed ones pass through silently.

The same plugin covers all four apps at once, no separate install needed per app. This guide covers installation for Microsoft Edge. For Chrome, see the Chrome integration guide. For Firefox, see the Firefox integration guide.

Prerequisites

• An active AccuKnox account with AI Security enabled
• Access to the Integrations section (to generate a token)
• Microsoft Edge browser

Step 1: Create a new integration

Go to AI Security > Integrations and click New Integration.

Fill in the following fields:

Field	Value
Integration Name	Any descriptive name (e.g. edge-browser-prod)
Tags	Add relevant tags (e.g. llmguard)

Click Add to save.

Step 2: Copy your token

After saving, the token is displayed once in a confirmation modal.

Save your token now

This token will not be shown again after you close the modal. Copy it immediately and store it in a secure location such as a password manager.

Step 3: Download and install the browser plugin

Download the plugin, download the ZIP file from Google Drive, and extract it to a folder on your machine.

To install in Edge:

1. Go to edge://extensions
2. Enable Developer mode (toggle in the left sidebar)
3. Click Load unpacked and select the extracted plugin folder

Step 4: Configure the extension

Click the AccuKnox icon in your browser toolbar and open Settings.

Fill in the following field:

Field	Value
API Key	Paste the token from Step 2

Click Save Settings, then Test Connection.

Step 5: Verify the connection

Once connected, the extension popup shows a green status dot next to AccuKnox Prompt Firewall along with a running count of scanned, allowed, blocked, and warned prompts. Click View Logs to see the full request log.

Open ChatGPT, Claude, GitHub Copilot, or Gemini and send a test prompt. Depending on your active policy, you will see one of the following:

Request log

All intercepted prompts and responses are visible under AI Security > Request Log.

Each entry shows:

• Timestamp
• Direction (PROMPT or RESPONSE)
• Verdict (ALLOW or BLOCK)
• Content preview
• Latency

You can filter by verdict or direction, search by content or reason, and export the full log as JSON.