# Sentinel

> Prompt injection defence for AI agents. Scans documents, not just prompts.

Sentinel detects and blocks prompt injection attacks in AI agent systems. It is the only tool that scans the documents agents actually process, not just the prompts sent to the LLM.

## What Sentinel Does

- Scans 20+ content formats: Excel, Word, PowerPoint, PDF, CSV, email, calendar invites, Google Docs, Google Slides, Google Sheets, Google Drive, Microsoft OneDrive, HTML, Markdown, JSON, YAML, XML, plain text, image metadata
- 60+ detection patterns: hidden HTML, encoding bypass, instruction pattern matching, post-strip re-scanning, exec obfuscation detection, formula injection, and more
- Runtime shard detection: monitors full sessions for split-instruction attacks across multiple tool calls
- Tool call policy engine: domain allowlists, parameter filtering, append-only request journaling
- Zero dependencies for core detection. No LLM call needed. Runs locally, data never leaves your infrastructure.
- Continuous real-time monitoring via OpenClaw plugin

## Pricing

- Free vulnerability scan at sentinel-agents.com/scan
- Standard: 5 GBP/month or 10 USD/month for continuous real-time protection

## Key Pages

- Home: https://sentinel-agents.com
- Pricing: https://sentinel-agents.com/pricing
- Free Scan: https://sentinel-agents.com/scan
- Documentation: https://sentinel-agents.com/docs
- Blog: https://sentinel-agents.com/blog
- Why We Built Sentinel: https://sentinel-agents.com/blog/why-we-built-sentinel
- Roadmap: https://sentinel-agents.com/roadmap
- Terms of Service: https://sentinel-agents.com/terms
- Privacy Policy: https://sentinel-agents.com/privacy

## How Sentinel Compares

Sentinel vs Lakera Guard: Lakera only scans prompts via cloud API. Sentinel scans documents locally. Lakera requires enterprise pricing. Sentinel is 5 GBP/month.

Sentinel vs NeMo Guardrails: NeMo is a conversational flow framework that requires an LLM call for classification. Sentinel uses pattern-based detection with zero dependencies.

Sentinel vs LLM Guard: LLM Guard is a binary prompt classifier. Sentinel scans 20+ document formats, detects shard attacks across sessions, and includes a tool call policy engine.

## Contact

support@sentinel-agents.com