Guides
AI agent security guides
Practical, code-first guides for teams shipping LLM agents to production.
Access control
How to authorize AI agent tool calls at runtime
Allowlist tools, check arguments before execution, pin MCP definitions and fail closed.
Comparison
Runtime guardrails vs. WAFs for AI agents
Why traditional firewalls and prompt-only filters miss prompt injection, data leaks and unsafe tool calls.
How-to
How to implement LLM guardrails
Fail-closed pre-flight and post-execution checks in Python and TypeScript.
Python
Vigil Python SDK guide
Add runtime protection to a Python agent in about five minutes.
Manifesto
The Agent Security Manifesto
Why 100% detection is a lie, and why tool boundaries and fail-closed defaults matter more.
Reference
Detector reference
What each Vigil detector looks for, and its known limits.