Overview & Architecture on The Mandrel Project

What's a Mandrel?

Mon, 01 Jan 0001 00:00:00 +0000

The Mandrel Project is a governance system for autonomous AI agents. An agent is bound to a declarative contract — the Mandrel Spec — that states its identity, what it is allowed to do, and the limits it operates under; a proxy called the Collet sits in front of it and enforces that contract on every turn. An action outside the contract does not happen.

The name comes from the machine shop, where a mandrel is the shaft that gives a workpiece a true axis while it is cut. The rest of the vocabulary follows from it, and this page explains why that vocabulary describes the problem better than a security vendor’s would.

Architecture

Mon, 01 Jan 0001 00:00:00 +0000

The Mandrel Project provides a deterministic governance layer for stochastic AI agents. It is designed to “clamp” amorphous agent behaviors to a rigid, machine-readable contract.

Design Philosophy: Engineering the Amorphous

In high-precision manufacturing, a Mandrel provides internal support and defines the shape of a workpiece. Similarly, the Mandrel Project provides the deterministic axis for risk mitigation in agentic systems:

The Mandrel Spec: A declarative contract defining an agent’s structural boundaries and risk envelope.
The Collet (Sidecar): A high-performance proxy that “clamps” the agent to the spec, ensuring it never slips from security, financial, or data-sovereignty constraints.
The Gantry: An automated capability for Agent Lifecycle Management, including sidecar injection and SPIFFE identity provisioning.
The Metrology Lab: The Promotion Authority where agents are certified and signed against platform policies and “Golden Datasets” to quantify performance risk. The role it fills is defined as a portable contract, so a deployment can satisfy it with the managed service, a self-hosted lab, or a CI pipeline holding a signing key — see the metrology contract.

Core Architectural Pillars

1. Runtime Mesh (Orchestrator-Specialist)

Mandrel utilizes a decoupled pattern enforced by the Collet sidecar:

Agent Workload Maturity Tiers

Mon, 01 Jan 0001 00:00:00 +0000

This guide defines the actionable technical criteria required to advance an agent workload through Mandrel’s four operational maturity tiers.

Maturity Alignment: While Enterprise Governance Maturity measures an organization’s overall program maturity from Shadow AI to a unified mesh, Workload Maturity Tiers govern the runtime promotion criteria for individual agent deployments. A mature organization enforces these automated workload gates within its CI/CD and runtime infrastructure.

Tier Summary

Tier	Posture	Identity & Access	Governance Gate	Target Capabilities
Tier 1: Sandbox	Private / Prototype	Anonymous / Dev Token	Static schema check	Read-only MCP resources
Tier 2: Trusted	Verified Workload	SPIFFE SVID + OIDC	`mandrel-cli validate`	Mandatory HITL (`challenge-always`)
Tier 3: Federated	Discoverable Mesh	Signed Realized Spec	Metrology Lab promotion	Intent classification & capability discovery
Tier 4: Autonomous	Hardened Runtime	SPIFFE Federation + DPoP	Continuous Conformance	Dynamic Client Registration & CAE revocation

Tier 1: Sandbox (Private Prototyping)

The entry point for rapid agent prototyping and local development.

Open Source Roadmap

Mon, 01 Jan 0001 00:00:00 +0000

The Mandrel open source ecosystem is planned in three phases: Core Foundation (current), Mesh Integration and Extensibility, and High-Assurance Tooling. Each phase covers The Mandrel Spec, the Mandrel CLI, and the Collet runtime enforcement engine.

Phase 1: Core Foundation (Current)

Focus: Establishing the deterministic contract and basic enforcement.

Mandrel Spec v1.0: Stable schema for identity, network boundaries, and threshold-based governance matrices.
Mandrel CLI: Static validator and declarative artifact generator for Kubernetes manifests and Open Policy Agent (OPA) policies.
Mandrel Collet (MVP): Local sidecar enforcement proxy for turn-level gRPC delegation, hop decrements, and basic proxying.
Chips Standard: Baseline format for Chip rate tables and per-model economic cost multipliers.

Phase 2: Mesh Integration & Extensibility

Focus: Enabling developer velocity, dynamic mesh topologies, and richer tool proxying.