AWARE is the open-source security control plane for autonomous AI agents. Cryptographic identity. Real-time policy enforcement. Bio-inspired coordination. Self-documenting audit. Compliance is one vertical you can build on it. The framework does much more.
Good CISO was founded by Alvin Chang — formerly Director of Enterprise Security Architecture at BT, where he led the team securing UK critical national infrastructure. The team co-authors the Cloud Security Alliance AI Controls Matrix, defining the global standard for AI agent governance. AWARE is Apache 2.0, 175+ tests passing, production-tested by UK regulated firms in finance and payments. This is open-source infrastructure for the autonomous agent era — not another closed compliance tool.
AWARE is the infrastructure layer every autonomous agent system needs. Five primitives, designed to compose.
RSA-2048 + SHA-256. Every action signed, every signature verified. T0–T4 constraint — invalid or missing signature blocks the action. No shared secrets, no service accounts, no override path.
Actions evaluated before execution, not after. T0–T4 constraint levels let you define autonomy per agent or per agent class.
Pheromone routing for multi-agent orchestration. Trust-weighted, security-aware, fail-safe by design.
Raft consensus for automatic revocation cascade. Compromise one agent — the network responds.
Every decision, every context access, every escalation. Exportable to any SIEM. Auditor-ready out of the box.
AWARE's identity model is a non-human identity (NHI) framework with cryptographic verification on every action. The T0–T4 constraint enforces signing at the most restrictive tier of the autonomy model. There is no override path.
// Real AWARE setup — register an agent and mint it a short-lived JWT
const IdentityProvider = require('aware/src/agents/identity-provider');
const AgentRegistry = require('aware/src/agents/registry');
const registry = new AgentRegistry();
const idp = new IdentityProvider({ secretKey: process.env.AWARE_SECRET, issuer: 'aware-ca' });
await registry.register({ agentId: 'agent-007', type: 'researcher', capabilities: ['web_search'] });
const { token } = idp.issueToken('agent-007');
Each agent gets an RSA-2048 keypair (SPKI public, PKCS8 private, both PEM). Every action is signed with RSA-SHA256 against a canonical payload — actionId|agentId|action|timestamp|paramsJson. Verification is signature-checked against the public key. Invalid or missing signatures block the action before it executes.
AWARE integrates with your agents in minutes, not months. Drop-in SDKs for Python, TypeScript, and Go.
Install the AWARE SDK. Wrap any agent with one line of code. Identity bootstrapped on first action.
Set your T0–T4 autonomy policy. AWARE evaluates every action against it — approved, denied, or escalated.
Every decision, every escalation, every context access is logged. Stream to your SIEM or query the audit API.
AWARE's compliance vertical maps the framework's primitives to the standards your regulators care about. Pre-built mappings, ready to deploy. Or build your own vertical on the framework.
"We went from zero documentation to ISO 27001 in 9 weeks. AWARE mapped our SaaS controls and produced the evidence pack the auditor signed off on. No external consultant."
"Our FCA compliance review used to take 6 weeks. AWARE does it in hours. We pass SYSC audits faster and the documentation is consistent every quarter."
AWARE is an open-source security control plane for autonomous AI agents. It provides cryptographic identity, real-time policy enforcement, bio-inspired coordination, and self-documenting audit — the infrastructure layer every autonomous agent system needs. Apache 2.0.
Traditional security tools — SIEM, IAM, EDR — were designed for human operators. AWARE was designed for autonomous agents. It provides per-agent identity, real-time action evaluation (not after-the-fact detection), and a kill switch with Raft consensus — primitives that simply don't exist in legacy security stacks.
AWARE defines T0–T4 constraint levels. T0: fully human-controlled, every action approved. T1: human-supervised, exceptions escalated. T2: bounded autonomy, irreversible actions escalated. T3: conditional autonomy with human-in-the-loop for high-stakes actions. T4: full autonomy with cryptographic identity and self-audit. You define the policy per agent or per agent class.
Yes. AWARE's compliance vertical maps the framework's primitives to CSA AI Controls Matrix, NIST AI RMF, ISO 27001, DORA, and OWASP LLM Top 10. Pre-built mappings, ready to deploy. The compliance vertical is one of many you can build on the framework.
Apache 2.0. Use it commercially, modify it, redistribute it. The only requirement is preserving the license and copyright notices. Good CISO retains copyright on the framework; the license grants broad reuse rights.
git clone https://github.com/GoodCISO/aware and then npm install. Wrap any agent with the AWARE wrapper. Set your T0–T4 policy. AWARE handles identity, policy evaluation, and audit. Most teams have a working integration in under an hour.
Open-source. Apache 2.0. 175+ tests. Production-tested. Fork it, deploy it, govern your agents.