Panguard AI secures AI agents through three pillars: ATR (the open standard for agent threat rules), Threat Cloud (collective immunity network), and Guard (the enforcement engine with skill audit and auto-response).
Platform
Three pillars. One mission:
secure every AI agent.
ATR defines the standard. Threat Cloud builds collective immunity. Guard enforces it all. One command to deploy -- works with Claude Code, Cursor, OpenClaw, and any AI agent.
Protected
3 endpoints active
847
Blocked
99.9%
Uptime
0
Alerts
ATR -- The Standard
StandardThe first open standard for AI agent threat detection.
ATR (Agent Threat Rules) is the first open standard for detecting AI agent threats. YAML-based rules across 10 categories cover prompt injection, tool poisoning, skill compromise, unauthorized access, data exfiltration, and more. ATR gives the security community a shared language for AI agent threats. Rules are open-source, human-readable, and machine-enforceable.
Learn more about ATR -- The StandardKey Capabilities
- Rules across 10 threat categories, growing continuously
- YAML-based, human-readable rule format
- Covers prompt injection, tool poisoning, skill compromise, data exfiltration
- Open-source -- community-contributed and reviewed
- Machine-enforceable by Guard and any compatible engine
- Versioned rule lifecycle: draft, experimental, stable, deprecated
Threat Cloud -- The Network
NetworkCollective immunity. Every install makes everyone safer.
Threat Cloud is a self-hosted collective intelligence network. Every Panguard install contributes anonymized threat signals; the pipeline auto-generates ATR rules from real-world attacks. 11 threat intel sources, 5,146 validated IoC records, and 808 community-promoted rules -- all synced every hour. The more nodes participate, the stronger everyone's defense.
Learn more about Threat Cloud -- The NetworkKey Capabilities
- 11 threat intel sources with 5,146 validated records
- Auto-generates ATR rules from collective data
- 808 community-promoted rules, synced every 1h
- Self-hosted -- your data never leaves your infrastructure
- Anomaly signals from Guard and Scan deployments worldwide
- Confidence scoring and rule lifecycle management
Guard -- The Engine
EngineSkill audit + known threat blocking + auto-response.
Panguard Guard is the enforcement engine. A 4-agent pipeline (Detect, Analyze, Respond, Report) processes AI agent events through 652 ATR detection rules. Built-in Skill Auditor runs 8 checks before any AI skill is installed. Three response modules auto-block IPs, kill processes, and quarantine files.
Learn more about Guard -- The EngineKey Capabilities
- 4-agent AI pipeline: Detect, Analyze, Respond, Report
- 652 ATR detection rules
- Skill Auditor: 8-layer pre-install security gate
- 3 auto-response modules: IP Blocker, Process Killer, File Quarantine
- Works with Claude Code, Claude Desktop, Cursor, OpenClaw, Codex, WorkBuddy, NemoClaw, ArkClaw, Windsurf, QClaw, Cline, VS Code Copilot, Zed, Gemini CLI, Continue, Roo Code, and any MCP-compatible AI agent
- Supports Linux, macOS, Windows, Docker, Kubernetes
Ready to secure your AI agents?
One command to deploy. Works with Claude Code, Cursor, OpenClaw, WorkBuddy, and any AI agent setup. 650+ ATR detection rules. Completely free.