Introduction

MDWR (Memory-Driven Workflow Runtime) combines workflow orchestration with operational decision memory to create workflows that remember, explain, and learn from every decision.

What is MDWR?

MDWR enables developers to build workflows that:

• Remember every decision and why it was made
• Explain decisions automatically (compile-time enforced)
• Learn from past decisions and outcomes
• Replay with different policies or memory-informed insights

Core Concepts

1. Workflow

A workflow is a sequence of steps that execute in order. Each step can make decisions based on input, state, and historical memory.

const workflow = Workflow.create("my_workflow_v1")
  .step("step1", handler1, true)
  .step("step2", handler2, true);

2. Step

A step is an executable unit that can make decisions. Steps that require decisions must return:

{
  decision: string;
  reason: string;
  confidence?: number;
  metadata?: object;
}

3. Execution

An execution is a single run of a workflow. You can pause, resume, and replay executions.

const execution = await workflow.execute({ userId: "u123" });
await execution.replay({ mode: "policy-aware" });

4. Memory

Memory stores every decision with context, allowing workflows to learn from history.

const insights = await ctx.memory.similar({
  situation: "my_decision",
  context: ctx.input
});

5. Events

Events signal workflow lifecycle and step completion, enabling event-driven architectures.

Architecture

MDWR uses:

• Motia as the backend runtime (event-driven, long-running)
• Mem0 as the memory layer (decision storage and similarity search)
• Redis for local memory storage (via Docker)

Developer SDK → MDWR Control Plane → Motia Runtime + Mem0/Redis

Next Steps

• Installation - Set up MDWR
• Quick Start - Build your first workflow
• Guides - Learn advanced features