Grimdall

About

Current AI security solutions typically focus on scanning code or managing secrets, missing the critical activity of agents during execution. Grimdall addresses this by intercepting every runtime tool call, including shell commands, file access, API interactions, and deployments. It enforces your security policies instantly, capable of blocking destructive actions, concealing sensitive data, or mandating human authorization. All activity is recorded in a verifiable audit log where each entry is cryptographically linked; altering even a single character will break the chain and trigger a verification failure. Setup requires just one command: npx grimdall init --hooks. The core is open source, operates with zero configuration, and functions completely offline.