Documentation

Welcome to the Mandrel Project documentation. Here you will find conceptual guides, architecture deep-dives, normative specifications, and technical references for our zero-trust agent governance ecosystem.

Role-Based Pathways

Select a reading path based on your role and operational focus:

• Agent Developers:
  Start with Getting Started to build your first spec, explore the Core Schema, and reference the Mandrel CLI Reference and Diagnostic Codes.
• Platform & Security Engineers:
  Review the Architecture, configure Identity & Security, deploy the Metrology Lab, and configure Egress Governance.
• Enterprise Evaluators & CISOs:
  Read the SAIF Alignment White Paper, assess the Enterprise Governance Maturity Framework, and follow the Workload Maturity Guide.

Documentation Directory

1. Overview & Architecture

• What’s a Mandrel?: Why precision governance borrows a machinist’s vocabulary.
• Architecture: How the Collet, Gantry, and Metrology Lab interact and where governance is enforced.
• Workload Maturity Tiers: Actionable checklist for advancing workloads from Tier 1 (Sandbox) to Tier 4 (Autonomous).
• Open Source Roadmap: Long-term development vision for the open-source governance core.

2. Guides & Tutorials

• Getting Started: Deploy your first governed agent in four steps: author, validate, test locally, and deploy to Kubernetes.
• Deployment Variants: Universal governance patterns across Kubernetes, Serverless, and Managed Agent Platforms.
• Egress Governance Architecture: Outbound tool proxying, Envoy transparent interception, and Collet fallback modes.
• Kong Gateway & Mesh Integration: Compile specs directly into declarative Kong AI Gateway policies and Mesh traffic permissions.

3. The Mandrel Specification

• Core Schema: Formal YAML schema for the MandrelSpec contract.
• Identity & Security: Compound Identity (SPIFFE + Nested JWTs), Continuous Access Evaluation (CAE), and Proof of Intent.
• Orchestration & Delegation: Task routing, max-hop budgets, and dynamic agent delegation.
• Extensions Reference: Governance controls, network allowlists, and the plugin extension block.
• PlatformPolicy Schema: Platform-scoped governance ceilings and parameter baking rules.
• ChipRateTable Schema: Per-model economic multipliers and Chip pegs.
• Metrology Contract: The normative contract defining promotion gate rules and signature verification.

4. Technical Reference

• Mandrel Collet: High-performance enforcement proxy sidecar and enforcement layers.
• Mandrel CLI: Command-line utility for validation, promotion, and artifact generation.
• Diagnostic Code Index: Complete listing of static validation error codes (E-codes, W-codes) and promotion codes (L-codes).
• Mandrel Gantry: Kubernetes operator for platform configuration and sidecar injection.
• Metrology Lab: Promotion gate service, credential issuer, and reference data host.

5. White Papers

• SAIF Alignment: Mandrel as the direct enforcement engine for Google’s Secure AI Framework.
• Governing Late-Binding Agent Harnesses: Pairing dynamic, network-enabled agent harnesses with deterministic governance.
• Enterprise Governance Maturity Framework: Evolving from Shadow AI to a unified, self-service governed agent mesh.
• Established Ecosystems: Introducing governance controls into existing, unmanaged agent environments.

6. Glossary

• Glossary: Canonical vocabulary and domain-specific definitions for the Mandrel Project.