How to Use Guardino Sentinel: The Essential Security and DLP Firewall for Generative AI

TL;DR: Guardino Sentinel is a real-time data loss prevention (DLP) firewall and browser extension for generative AI. It masks API keys, passwords, and private company data in local memory before ChatGPT or Claude can store them, while neutralizing prompt injection attacks at zero latency cost.

What is Guardino Sentinel?

Guardino Sentinel is an AI-native data loss prevention (DLP) firewall and security proxy designed to intercept, sanitize, and neutralize sensitive data leaks and prompt injection attacks before prompts reach external foundation models.

When engineers and knowledge workers query public artificial intelligence platforms like ChatGPT or Claude, they frequently paste production log snippets, database connection strings, API keys, or confidential customer communications. Guardino Sentinel operates as an in-flight security shield. Running client-side within the browser or developer environment, it detects confidential tokens, applies immediate cryptographic masking, and blocks jailbreak vectors before network packets hit AI vendor inference servers.

To see how network-layer DNS filtering complements endpoint AI security, read our complete Guardino AI protection guide and check our pricing plans.

The Hidden Attack Surface of Generative AI

Public artificial intelligence models offer tremendous productivity, but introduce critical enterprise compliance and security hazards:

1. Unintended Training Data Ingestion: Cloud models frequently ingest user inputs for reinforcement learning and model fine-tuning. Pasting proprietary source code or private customer details exposes intellectual property.
2. Hardcoded Credential Exfiltration: Developers frequently paste shell traces or .env files into AI chats to debug errors. AWS secrets, GitHub tokens, and private SSH keys are instantly exposed.
3. Indirect Prompt Injections: When an AI agent browses untrusted web content or reads email attachments, hidden adversarial instructions can override the system prompt and compromise sensitive corporate tools.

Capability	Guardino Sentinel	Plain Chatbot (Unprotected)	Enterprise Chatbot Contract	Static Code Linter
Real-Time In-Flight Redaction	Yes (Local Memory)	No	No	No (Pre-commit only)
OWASP LLM Top 10 Injection Defense	Yes (Active)	Vendor Filter Only	Vendor Filter Only	No
Ghost Mode (Client-Side Privacy)	Yes (100% Local)	No	Cloud Storage	Local Only
Canary Trap Leak Telemetry	Yes (Cryptographic Trace)	No	No	No
Agentic MCP Integration (Claude/Cursor)	Yes (Native)	No	No	No
Tamper-Evident Security Dashboard	Yes (Encrypted)	Standard History	Admin Logs Only	CI Reports

Step-by-Step: How to Use Guardino Sentinel

Setting up Guardino Sentinel across your daily AI workflows takes less than two minutes:

Step 1: Install the Browser Extension

1. Navigate to the official Chrome Web Store or developer portal download page.
2. Click Add to Chrome (or Edge / Brave / Arc).
3. Pin the Guardino Sentinel shield icon to your browser toolbar for instant inspection telemetry.
4. Open the extension popup, log into your Guardino account, and verify that the green Shield Active badge illuminates.

Step 2: Activating Ghost Mode for Total Privacy

When Ghost Mode is enabled, all prompt parsing and secret sanitization run entirely within your local browser memory sandbox.

• Whenever you paste code into ChatGPT, Claude, or Gemini, Sentinel inspects text fields in real time.
• If high-entropy strings, API tokens (sk-ant-..., ghp_..., AKIA...), or database URIs are detected, Sentinel swaps them dynamically with a deterministic pseudonym: [REDACTED_BY_SENTINEL].
• The external AI model responds accurately to the abstracted code without ever seeing your actual infrastructure secrets.

Step 3: Integrating with Developer AI Agents via MCP

For software engineers utilizing agentic tools such as Claude Code, Cursor, or Cline, Guardino Sentinel provides an official Model Context Protocol (MCP) server package: @guardino/mcp.

1. Generate an API key in your setup portal.
2. Configure the Sentinel MCP server in your local environment configuration:

{
  "mcpServers": {
    "guardino-sentinel": {
      "command": "npx",
      "args": ["-y", "@guardino/mcp@latest"],
      "env": {
        "GUARDINO_API_KEY": "gdo_live_your_api_key_here"
      }
    }
  }
}

3. Your AI coding assistant can now verify DNS shield postures, switch Mind Modes, and sanitize prompt payloads automatically.

Reviewing Security Events on Your Dashboard

All security detections appear chronologically in your central console at /dashboard/sentinel:

• Blocked Injections: Displays attempts to hijack system instructions with rule categorizations (e.g., LLM-01: Prompt Injection).
• Redacted Secrets: Shows the timestamp, platform surface (e.g., ChatGPT or Claude), and detected credential classification (AWS, Stripe, OpenAI, SSH) without revealing the underlying secret.
• Privacy Audit Log: Generates clean, tamper-evident compliance records suitable for internal data governance, security audits, and GDPR compliance reviews.

Frequently Asked Questions

What is Guardino Sentinel and how does it protect my data?

Guardino Sentinel is an AI security proxy and browser extension that inspects prompts in real time. It redacts API keys, credentials, and sensitive personal data locally before tokens are transmitted to external AI providers.

Does Guardino Sentinel send my prompts to an external server?

No. With Ghost Mode enabled, prompt inspection and secret masking execute entirely inside your local browser memory. Your raw proprietary code and confidential prompts never leave your device.

Which AI platforms are currently protected by Guardino Sentinel?

Guardino Sentinel natively protects web sessions across ChatGPT, Anthropic Claude, Google Gemini, DeepSeek, and Microsoft Copilot, as well as developer environments using Claude Code and Cursor.

How does Guardino Sentinel defend against prompt injection attacks?

Sentinel evaluates incoming prompts and untrusted inputs against OWASP LLM Top 10 heuristic rules, neutralizing indirect injections, jailbreak payloads, and system instruction overrides before they trigger execution.

What is the Canary Trap feature in Guardino Sentinel?

The Canary Trap injects tamper-evident token signatures into test queries. If a third-party AI provider inadvertently trains on or leaks your data, the canary token alerts your dashboard immediately.

Secure Your Artificial Intelligence Workflows Today

Generative AI offers massive productivity, but data security cannot be compromised. With Guardino Sentinel, your developers and teams leverage artificial intelligence fearlessly without leaking corporate IP.

Activate Guardino Sentinel today in your dashboard at https://my.guardino.ai/dashboard/sentinel and explore our easy setup instructions.