Skip to content

PolicyShield

Declarative firewall for AI agent tool calls.

PolicyShield lets you define YAML rules that control what tools an AI agent can use, when, and how — without modifying agent code.

Key Features

  • 🛡️ Declarative rules — YAML-based, no code changes needed
  • 🔗 Chain rules — Temporal conditions for multi-step attack detection
  • 🔍 PII detection — Built-in redaction for sensitive data
  • 🚨 Kill switch — policyshield kill for instant emergency stop
  • 🍯 Honeypot tools — Decoy tools that catch prompt injection
  • 🩺 Doctor — 10-check health scanner with A–F grading
  • ⚙️ Auto-rules — policyshield generate-rules --from-openclaw
  • ✅ Approval flows — Human-in-the-loop for risky operations
  • 📊 Tracing — Full audit trail of every tool call
  • 🔄 Replay & Simulation — Re-run traces against new rules
  • 🤖 AI Rule Writer — Generate rules from natural language
  • 🌐 HTTP Server — Framework-agnostic REST API for tool call policy enforcement
  • 🔌 OpenClaw Plugin — Native plugin with before/after hooks
  • 🔗 Integrations — LangChain, CrewAI
  • 🧪 Testing — Validate rules before deployment
  • 🚀 CLI — Scaffold, validate, lint, test, serve from the command line

Quick Start

pip install policyshield

# Scaffold a new project with secure defaults
policyshield init --preset secure --no-interactive

# Check your security posture
policyshield doctor

# Validate your rules
policyshield validate policies/

# Start the HTTP server
pip install "policyshield[server]"
policyshield server --rules policies/rules.yaml --port 8100

How It Works

# policies/rules.yaml
shield_name: my-policy
version: 1

rules:
  - id: block-file-delete
    when:
      tool: delete_file
    then: block
    severity: high
    message: "File deletion is not allowed."

from policyshield.shield.engine import ShieldEngine

engine = ShieldEngine(rules="policies/rules.yaml")
result = engine.check("delete_file", {"path": "/data"})
print(result.verdict)  # Verdict.BLOCK

Next Steps