Engineering Amorphous AI

The Mandrel Project provides a declarative contract and enforcement ecosystem for autonomous agent governance.

Why Mandrel?

As AI agents move from “chatbots” to “autonomous actors,” the need for rigorous, machine-readable governance becomes critical. Traditional security models are ill-suited for the fluid, non-linear execution patterns of agentic systems.

Mandrel bridges this gap by providing Zero-Trust Enforcement for AI agents, ensuring they operate strictly within their defined operational boundaries and risk envelopes.

Design Philosophy

1. Declarative Governance

We believe governance should be a contract, not an afterthought. Every Mandrel agent is bound to a Mandrel Spec that defines its identity, capabilities, and constraints.

2. Zero-Trust by Design

Every “turn” in an agentic workflow is intercepted and verified. No agent is trusted implicitly; every action must be justified against its governing matrix.

3. High-Performance Enforcement

Security shouldn’t be a bottleneck. Our enforcement proxy, the Mandrel Collet, is engineered for sub-millisecond overhead to maintain real-time agent responsiveness.

Component Ecosystem

Mandrel Spec

The declarative contract for autonomous agent governance. Language-neutral schemas for operational boundaries.

Read the spec

Mandrel Collet

High-performance enforcement proxy (sidecar) that “clamps” agents to their specifications.

Read the docs

Mandrel CLI

Governance tooling for validating specs and generating deployment artifacts.

Read the docs

Mandrel Gantry

Kubernetes Operator for agent lifecycle management and SPIFFE identity provisioning. Coming soon.

Learn more

Metrology Lab

Attestation registry and semantic discovery gateway for the governed agent mesh. Coming soon.

Learn more