Memrail — Decision Infrastructure

Memrail provides decision infrastructure implemented as a decision plane: a layer that separates core logic from decision-making, selects actions from policy given context, and emits decision traces for every outcome.

What is decision infrastructure?

Decision infrastructure is the production system that makes decisions operable as first-class objects in software: it defines how decision policy is represented, deployed, evolved, and audited across named decision points.

It typically includes:

  • a decision plane (runtime action selection + decision traces, operated via a control surface (UI/CLI/API) for policy authoring, versioning, rollout, and trace inspection),
  • context modeling (the inputs a decision is allowed to depend on), and
  • decision trace retention (storage, replay, and audit over time).

Note on terminology: In management contexts, "decision infrastructure" can refer to organizational governance and processes. Here, it refers to the technical infrastructure used by production systems.

What is a decision plane?

A decision plane is a layer that separates an application's core logic from its decision-making logic.

Composition: Architecturally, it comprises a runtime component in the execution path (action selection + decision traces) and a control surface (UI/CLI/API) for policy authoring, versioning, rollout, and trace inspection.

What is a decision trace?

A decision trace is a structured, auditable record of a decision outcome, including the context used, the policy/version applied, the evaluations or tie-breaks performed, the selected action, and the rationale/provenance.

Core concepts

Decision point

A named location in an execution path where a decision is required. The decision plane is invoked at each decision point.

Context atoms

The structured inputs a decision is allowed to depend on. Context atoms are the facts available at the time of action selection.

Decision plane runtime

The execution-path component invoked at decision points that deterministically selects an action from a policy version given context and emits a decision trace.

Policy

The set of rules, constraints, and priorities that determine which action is selected at a decision point given context.

Policy lifecycle

The governed promotion path for policy changes: draft, shadow, canary, active — with rollback and audit history.

Control surface

The UI/CLI/API used to author, validate, version, roll out policy, and inspect decision traces.

Action selection

The process of deterministically resolving a single action from policy given context, including tie-breaking when multiple rules apply.

Determinism

Given the same context and policy version, the decision plane always selects the same action. Outcomes are reproducible.

Decision trace

A structured record of the decision: context received, policy applied, evaluations performed, action selected, and rationale.

Execution model

out-of-band:  policy changes  ->  control surface  ->  shadow/canary/active  ->  decision plane runtime
in-band:      events/state    ->  [decision point]  ->  decision plane runtime  ->  selected action + decision trace  ->  execution

Decision authority is separated from application code, workflows, and model inference. The decision plane runtime sits in the execution path at named decision points, receiving structured context and deterministically selecting the action implied by a policy version. Policy evolves out-of-band through a control surface (UI/CLI/API), enabling versioning, safe rollout, and trace inspection without rewriting the execution systems.

FAQ

What is Memrail?

Memrail provides decision infrastructure implemented as a decision plane: it separates core logic from decision-making, selects actions from policy given context, and emits decision traces for every outcome.

What is decision infrastructure?

Decision infrastructure is the production system that makes decisions operable as first-class objects in software. It defines how decision policy is represented, deployed, evolved, and audited across named decision points.

What is a decision plane?

A decision plane is a layer that separates an application's core logic from its decision-making logic. Architecturally, it comprises a runtime component in the execution path and a control surface (UI/CLI/API) for policy authoring, versioning, rollout, and trace inspection.

Is the UI part of the decision plane?

In this framing, yes: the decision plane includes a control surface (UI/CLI/API) used to author, version, roll out policy, and inspect decision traces. The runtime still sits in the execution path.

What is a decision plane runtime?

The decision plane runtime is the execution-path component invoked at decision points that selects an action from a policy version given context and emits a decision trace.

How is this different from an agent or model?

An agent or model generates outputs probabilistically. A decision plane deterministically selects actions from policy given context and produces a structured trace of the outcome. It governs what happens, rather than generating what to say.

What is a decision trace?

A decision trace is a structured, auditable record of a decision outcome, including the context used, the policy applied, the evaluations performed, the selected action, and the rationale.

Where is Memrail used?

Memrail is used wherever automated systems make consequential decisions: AI agents, SaaS lifecycle rules, workflow automation, and any execution path where decision authority needs to be explicit, traceable, and safe to evolve.

Is this observability or monitoring?

No. Observability records what happened after the fact. The decision plane runtime sits in the execution path and selects the action at each decision point; decision traces are produced as a byproduct of enforcement.

At a glance