Edera
Hardened runtime isolation for AI workloads — per-container micro-VMs preventing lateral movement.
What it does
Hardened runtime isolation for AI and container workloads. Replaces shared Linux kernels with per-container micro-VMs, providing complete workload isolation. GPU workload isolation prevents data leakage between tenants. $20M total funding, M12 (Microsoft) led Series A. Performance within 5% of native containers.
Security relevance
Solves the sandboxing problem for AI agents — each agent runs in full isolation without access to host OS, file system, or other agents. Prevents privilege escalation, lateral movement, and container escapes. Research showed running agents in Edera can actually be faster than Docker while being significantly more secure.
When to use it
Deploy when running AI agents that generate or execute code, or when multi-tenant AI workloads share infrastructure. Requires Kubernetes expertise and infrastructure-level deployment. Essential for any production multi-agent system where agents have meaningful permissions.
OWASP coverage
Risks addressed — mapped to both OWASP Top 10 standards. 2 in LLM, 2 in Agentic.
The raw record
What Yuntona stores. Single source of truth — fork it on GitHub.
name: Edera slug: edera type: Mixed category: AI Guardrails & Firewalls url: https://edera.dev reviewed: 2026-04 added: 2026-04 updated: 2026-04 risks: llm: [LLM02, LLM07] asi: [ASI05, ASI08] complexity: Expert Required pricing: — audience: Builder lifecycle: [deploy] tags: [Container, Infrastructure, Isolation, Sandbox]