Docs/Predicate Secure

Predicate Secure

Drop-in security wrapper for AI agents. Adds authorization, verification, and audit to any agent framework in 3 lines of code.

What is predicate-secure?

predicate-secure is a drop-in security wrapper that adds authorization, verification, and audit capabilities to any AI agent framework. Instead of rewriting your agent code, you simply wrap your existing agent with SecureAgent and define a policy file.

Key Features

  • Pre-action authorization - Every action is checked against your policy before execution
  • Post-execution verification - Deterministic checks ensure the expected outcome occurred
  • Cryptographic audit - All decisions are logged with tamper-proof receipts
  • Zero refactoring - Works with your existing agent code

Supported Frameworks

FrameworkDetectionStatus
browser-useAutoFull support
PlaywrightAutoFull support
LangChainAutoFull support
PydanticAIAutoBasic support
OpenClawAutoFull support (TypeScript only)

How It Works

Your Agent Code
      │
      ▼
┌─────────────────┐
│  SecureAgent    │
│  ┌───────────┐  │
│  │  Policy   │◀─── Your rules (YAML)
│  │  Engine   │  │
│  └───────────┘  │
│  ┌───────────┐  │
│  │ Snapshot  │◀─── Before/after state
│  │  Engine   │  │
│  └───────────┘  │
│  ┌───────────┐  │
│  │  Audit    │◀─── Decision log
│  │   Log     │  │
│  └───────────┘  │
└─────────────────┘
      │
      ▼
  Execution

Execution Modes

ModeFail ClosedDescription
strictYesDeny unauthorized actions, halt on failure
permissiveNoLog but allow unauthorized actions
debugNoFull trace output for development
auditNoRecord all actions for compliance

Quick Links

GitHub Repositories

Previous
Policy Signing (Ed25519)
Next
Installation