Documentation
Make dangerous AI agent actions structurally impossible.
Security Layer is an agent security platform that enforces capability-based access control, taint tracking, action normalization, and deterministic rules — so dangerous actions are structurally impossible, not just detected.
Getting started
Learn more
Concepts
Understand taint levels, capabilities, and the security pipeline.
CLI commands
Full reference for every command.
Integrations
Protect Claude Code, Cursor, and any shell-based agent.
Guides
Practical guides for common workflows.