# Vigil Guard

> Enterprise AI Detection and Response (AIDR) platform. Runtime security
> for prompts, AI outputs, and autonomous agents. Real-time prompt injection
> detection, PII protection, content moderation, and policy enforcement
> across LLM applications.

Vigil Guard provides a single security control layer for employee AI usage,
AI-enabled applications, and autonomous agents. GDPR-compliant. Built for
security teams operating LLM deployments at enterprise scale.

Founder and CEO: Tomasz Bartel (20+ years cybersecurity, formerly CrowdStrike,
Cisco, IBM, Riverbed).

## Core pages

- [Platform](https://www.vigilguard.ai/platform/): AIDR platform capabilities, architecture, detection engines.
- [Pricing](https://www.vigilguard.ai/pricing/): Subscription tiers and enterprise plans.
- [Developers](https://www.vigilguard.ai/developers/): SDK, REST API, n8n nodes, integration guides.
- [Install](https://www.vigilguard.ai/install/): Deployment options (SaaS, self-hosted, hybrid).
- [News](https://www.vigilguard.ai/news/): Product releases, security research, AI threat intelligence.
- [Blog](https://www.vigilguard.ai/blog/): Long-form analysis on AI security, AIDR strategy, prompt injection defense, agent observability.
- [Blog RSS](https://www.vigilguard.ai/blog/rss.xml): RSS feed for new blog posts.
- [About](https://www.vigilguard.ai/about/): Company overview, founder, mission.
- [Contact](https://www.vigilguard.ai/contact/): Demo requests and enterprise inquiries.

## Detection capabilities

- Prompt injection detection (direct, indirect, agentic)
- PII detection and redaction
- Content moderation
- Scope drift detection (agent stays within defined mission)
- Agent context logging (full visibility into agent decisions)
- Dual-pass analysis for long prompts
- VGE PromptGuard v1g classifier (44M params, 0.981 F1 on direct injection)
- VGE CC-Guard: local security sidecar for Claude Code. Gates tool calls, scans outputs through VGE, BLOCK-default policy, JSONL audit log.

## Key concepts

- AIDR (AI Detection and Response): runtime security category for AI systems, analogous to EDR for endpoints and NDR for networks.
- Prompt injection: adversarial input designed to override LLM instructions, exfiltrate data, or hijack tool use.
- PII detection: identifying personally identifiable information in prompts and outputs before they leave the trust boundary.
- Agent observability: capturing full context of autonomous agent actions including tool calls, intermediate reasoning, and scope drift.
- Scope drift: when an autonomous agent operates outside its intended mission scope.

## Languages

- English (default): https://www.vigilguard.ai/
- Polish: https://www.vigilguard.ai/pl/

## Technical resources

- Open-weight detection model: https://huggingface.co/VigilGuard/vigil-llm-guard
- VGE CC-Guard install guide (PDF): https://www.vigilguard.ai/datasheets/claude-code-guard-install-guide-en.pdf
- VGE CC-Guard user guide (PDF): https://www.vigilguard.ai/datasheets/claude-code-guard-user-guide-en.pdf
- n8n community node: https://n8n.io (search VigilGuard)
- LinkedIn: https://www.linkedin.com/company/vigil-guard/
- GitHub: https://github.com/vigil-guard