Skip to content

Gen AI Browser Integration for Prompt Security (Claude)

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

Red banner - Prompt blocked

This guide covers installation for Chrome, Brave, and Edge (Chromium-based browsers). The same plugin works for both Claude and ChatGPT — no separate install needed.

Prerequisites

  • An active AccuKnox account with AI Security enabled
  • Access to the Integrations section (to generate a token)
  • A Chromium-based browser (Chrome, Brave, or Edge)

Step 1: Create a new integration

New Integration modal

Go to AI Security > Integrations and click New Integration.

Fill in the following fields:

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

Click Add to save.

Step 2: Copy your token

Token Created Successfully modal

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 Chrome plugin and extract the ZIP file to a folder on your machine.

To install in Chrome:

  1. Go to chrome://extensions
  2. Enable Developer mode (top-right toggle)
  3. Click Load unpacked and select the extracted plugin folder

Step 4: Configure the extension

AccuKnox Prompt Firewall Settings panel

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

Fill in the two fields:

Field Value
API Key Paste the token from Step 2

Click Save Settings, then Test Connection.

Step 5: Verify the connection

Extension popup showing connected status and scan counts

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 Claude 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.

Request Log table