AKIOS
Introducing EnforceCore: Runtime Enforcement for AI Agents
Open-source enforcement framework extracted from AKIOS. Policy engine, PII redaction, Merkle audit, and resource guards — in a single decorator. Apache 2.0.
Blog
Release notes, deep dives, and practical guides.
AKIOS v1.0.7: Audit Integrity, Workflow Validation, and 53 PII Patterns
Real Merkle proofs, PII fail-safe hardening, workflow validation, ablation benchmarking, audit stats/rotate, and 10 new digital identity patterns.
AKIOS v1.0.5: Rich TUI, Accessibility, and Critical Fixes
Rich terminal UI, environment detection, color-blind modes, new diagnostic commands, and a SIGTRAP fix.
AKIOS v1.0: The Security Cage is Here
The strongest open-source security foundation for AI agents. Run with confidence — agents can't leak data, burn budgets, or escape their sandbox.
Hardening Checklist for AKIOS 1.0
Baseline steps to run AKIOS safely in prod-like environments: kernel settings, audit sinks, policy signing, and CI guardrails.
Policy Schema Deep Dive (v1)
A walkthrough of the v1 policy schema with examples for filesystem, HTTP, LLM, and tool execution.
Secure Agent Communication: How AKIOS Isolates Agent Interactions
How AKIOS mediates every agent interaction through policy-gated communication, PII redaction, and cryptographic audit trails.
AKIOS Roadmap: 2026 and Beyond
The strategic direction for AKIOS Open Runtime. From usability improvements to ecosystem growth.
Introducing AKIOS: A Secure Runtime for Untrusted AI Agents
Why we built a dedicated runtime for local AI agents, and how it differs from traditional sandboxing.