Maudel is an AI engineering system that combines deterministic orchestration with an artifact traceability graph to reliably produce software artifacts from idea to deployment. It transforms AI from an unpredictable assistant into a governed engineering process.

The Deterministic Orchestrator is the governing control system of Maudel. It manages pipeline stage execution, task assignment, artifact validation, dependency resolution, quality gates, and retry/revise/escalate decisions. Unlike typical agent systems, agents do not decide when work is complete — the orchestrator evaluates readiness and advances or blocks the pipeline.

Maudel's pipeline progresses through six deterministic stages: Scaffold, Priming, Review, Generate Plan, Worktree, and Execute. Each stage has quality gates evaluated by the orchestrator, with clear inputs, outputs, and progression criteria — ensuring repeatable, auditable execution every time.

The Artifact Traceability Graph is a directed lineage graph connecting every artifact produced during the pipeline — from compass documents and architecture specs to code, tests, and deployment configurations. Each artifact is connected to its parent artifacts, forming a verifiable chain from intent to delivery.

The Pipeline Minder is Maudel's real-time quality supervision system. It uses 12 pluggable evaluators covering worktree isolation, git state, build awareness, input/output quality, scope compliance, and more. It provides severity-gated auto-remediation, cross-execution pattern detection, and a live dashboard with findings feed and end-of-run reports.

Maudel's Human-in-the-Loop system keeps humans as first-class participants at every critical decision point. It includes decision handling with toast notifications, configurable timeouts with auto-proceed options, steering message injection for live operator guidance, and full audit trail tracking for compliance and review.

Maudel determines project readiness using the artifact traceability graph. Validation rules ensure no orphan artifacts exist, all artifacts are linked to upstream intent, required downstream evidence is present, tests are mapped to acceptance criteria, and deployment artifacts are defined. This allows Maudel to compute traceability coverage across the entire project.