Skip to content

AEGIS Governance Decision SDK

Quantitative governance for engineering decisions.

AEGIS evaluates proposals through Bayesian confidence gates, complexity analysis, and utility functions — returning structured PROCEED / PAUSE / ESCALATE / HALT decisions with full audit trails.

Get Started Try the Advisor API Reference

Why AEGIS?

Problem AEGIS Solution
"Should we ship this?" is a gut call Six quantitative gates with Bayesian confidence
No audit trail for engineering decisions Every evaluation gets a unique decision_id and timestamp
Risk thresholds are tribal knowledge Configurable, frozen parameters with version-controlled schemas
AI agents act without governance MCP server, GitHub Action, and SDK for agent-native integration

How It Works

Proposal → [Risk] [Profit] [Novelty] [Complexity] [Quality] [Utility] → Decision
              ↓                                                            ↓
         Bayesian posterior                              PROCEED / PAUSE / HALT / ESCALATE
         P(delta >= 2 | data)                           + confidence + rationale + next steps

Every proposal is evaluated against six gates:

Gate Method Default Threshold
Risk Bayesian posterior > 0.95 confidence
Profit Bayesian posterior > 0.95 confidence
Novelty Logistic function >= 0.8
Complexity Normalized floor >= 0.5
Quality Min score + no zero subscores >= 0.7
Utility Lower confidence bound > 0

Quick Example

from aegis_governance import AegisConfig, PCWContext, PCWPhase, pcw_decide

config = AegisConfig.default()
evaluator = config.create_gate_evaluator()

decision = pcw_decide(
    PCWContext(
        agent_id="my-agent",
        session_id="session-1",
        phase=PCWPhase.PLAN,
        proposal_summary="Add Redis caching layer",
        estimated_impact="medium",
        risk_proposed=0.15,
        complexity_score=0.7,
    ),
    gate_evaluator=evaluator,
)

print(decision.status.value)  # "proceed"

curl -X POST https://api.aegis-governance.com/evaluate \
  -H "X-Api-Key: YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "proposal_summary": "Add Redis caching layer",
    "estimated_impact": "medium",
    "risk_proposed": 0.15,
    "complexity_score": 0.7
  }'

echo '{"proposal_summary":"Add caching","estimated_impact":"low"}' \
  | aegis evaluate

Integration Options

Method Best For Docs
REST API Any language, CI/CD pipelines Quickstart
Python SDK Python applications, scripts Quickstart
CLI Shell scripts, local evaluation Quickstart
MCP Server AI agent integration MCP Tools
GitHub Action PR governance gates Action

Key Features

  • Zero runtime dependencies — core SDK uses only Python stdlib
  • 3029 tests, ~94.8% coverage — battle-tested across Python 3.9-3.12
  • Post-quantum cryptography — Ed25519 + ML-DSA-44 hybrid signatures
  • Full audit trails — hash-chained, tamper-evident decision logs
  • Configurable thresholds — YAML-driven parameter management
  • Shadow mode — evaluate without affecting production decisions
  • Drift detection — KL-divergence monitoring for parameter drift