GRIMDALL SECURITY
Runtime security for AI agents.
Intercepting tool calls, blocking dangerous commands, auditing everything.
grimdall.site · github · @AniketVarshne
A runtime security layer that stops rogue AI agents from deleting production.
The core idea: agents don't get unrestricted access to your systems. Every tool call passes through an interception layer that validates intent, blocks dangerous operations, and produces a full audit trail. No model, no network calls, no API keys required to enforce the policy.
grimdall-os — Local-first runtime security layer. Intercepts tool calls, blocks dangerous commands, audits everything. TypeScript
HarnessGuard — Zero-cost, offline static analyzer that catches security and reliability flaws in AI agent orchestration code before it ever runs. Python
MemLens — Memory observability for LLM agents. Trace what they remember, why, and how it shapes every response. Python
- Agent runtime security — sandboxing, tool-call interception, blast-radius containment
- Static analysis — catching orchestration flaws before execution
- Memory & observability — understanding what agents retain and why
- Local-first architecture — security that doesn't phone home