HOME OF METHOD-DRIVEN ENGINEERING
MDE guides AI through the engineering work: interviewing you, modelling your requirements, creating a prototype, designing, building and verifying your application. You provide business knowledge, review the results and make the decisions.
Open source · MIT licensed · In active development
01 / SIX STEPS, LED BY AI
AI follows MDE's engineering method and produces the work at each stage. You answer its questions, review its proposals and ask for changes; AI updates the specifications and implementation. The goal and strategy determine which stages are needed; the sequence below explains a typical journey.
AI asks about the problem, the people involved and how work happens today, then captures goals, exceptions and business rules from your answers.
You review: the business understanding.
AI organizes entities, relationships, roles, use cases and rules into a structured model, then revises it in response to your feedback.
You review: what the application must do.
AI creates sample data and a prototype so you can try the workflows. It uses your feedback to correct misunderstandings before the build.
You review: how the work should feel.
AI proposes design and architecture, including the technology choices and engineering concerns relevant to your project.
You review: design and architecture.
AI builds against the reviewed specifications and agreed stack, updating the specifications when your feedback changes the requirements.
You review: the resulting workflows.
AI runs checks against the requirements, records results, failures and gaps, and addresses findings. You review the evidence and decide whether the outcome meets your needs.
You review: what has actually been demonstrated.
02 / TWO KINDS OF KNOWLEDGE
MDE keeps reusable engineering methods distinct from the structured requirements model for your application. Domain knowledge and industry patterns help inform that model.
REUSABLE / THE METHOD
Strategies guide the work. The metamodel defines specification concepts. Features and Targets bring reusable guidance for concerns such as validation, authorization, history and auditability.
Example: guidance for enforcing authorization across an application.
PROJECT-SPECIFIC / THE MODEL
AI captures your people, entities, relationships, use cases and business rules in a structured requirements model. It develops design and architecture from that model for your review.
Example: a receptionist may book an appointment only with a practitioner qualified for that service.
Both remain visible and editable. Ask AI to explain or revise a rule; review the updated model before that understanding becomes code.
03 / A SERVICE-BUSINESS EXAMPLE
Imagine a service clinic coordinating clients, services and practitioners. The goal is simple: let staff book a suitable appointment without double-booking anyone.
“We need to match each appointment to someone with the right skills, during their available hours.”
People: clients, reception staff and practitioners.
Records: services, availability and appointments.
Rules: required skills, full-duration availability and no overlapping bookings.
STEP-BY-STEP / MDE TUTORIAL
Follow a real business application from business understanding through requirements, design, architecture, implementation, verification and evidence. The tutorial uses actual MDE Workbench screenshots and shows where engineering knowledge enters the process.
Open the MDE tutorial →TRY & WATCH / DEMOS
Explore a small scheduling example, then open recorded MDE walkthroughs to see the engineering process.
Choose a service, practitioner, and time. See how explicit business rules govern which appointments can be booked.
Avery: assessments & follow-ups · 09:00–12:00
Jordan: follow-ups · 13:00–16:00
An appointment needs a service, a qualified practitioner, and a start time. Its full duration must fit within that practitioner's available hours. Two appointments for the same practitioner must not overlap.
DEMO ENVIRONMENT — Nothing here is real or saved. This example illustrates how MDE business rules can govern an application. Sample appointments disappear when the page reloads.
These archived recordings show an earlier version of MDE. Commands and screens may differ from the current release. GitHub access is required.
Open the existing demonstration recording from the MDE archive.
Open recording on GitHubExplore the archived business analysis recording alongside the original user story.
Open recording on GitHubEXPLORE / THE WORKBENCH
The Workbench makes your project visible: its specifications, running prototype, current goal, and verification evidence.
Explore the Workbench guideAn example of the project information MDE helps you explore.
VERIFY / EVIDENCE YOU CAN REVIEW
AI verifies the implementation against the specifications and records what was checked, the result and what remains unresolved. You use that evidence to assess whether the application is ready for its intended use.
AI identifies the rule or use case each check exercises. For the booking example: a practitioner cannot have overlapping appointments.
AI collects executed test results, workflow screenshots and relevant audit findings for your review. A planned or unrun check is not a passed check.
AI records failures, missing coverage and limitations that need a decision. You decide which limitations are acceptable; passing a demo alone does not establish production readiness.
This homepage shows illustrative examples, not a completed project's verification report. See how evidence fits into the tutorial →
CHOOSE / YOUR GOAL, YOUR RIGOR
Choose the level of rigor for AI's work: exploring an idea, building with guidance or verifying every stage.
AI makes the idea tangible with diagrams, sample data, and pages on the MDE runtime.
AI carries out the engineering work and surfaces gaps. You review its progress and decide what matters before release.
AI requests your approval of key choices, verifies each stage, and produces evidence for your release decision.
RESOURCES / THE MDE ECOSYSTEM
The method, the tools, and your project's specifications remain visible and editable.
Explore the MDE source, the engineering method, and the Workbench.
View repositoryBUSINESS MODELSExplore reusable business concepts and models across industries and common business functions.
Explore patternsDOCUMENTATIONUnderstand the concepts, write your business story, and begin your first project.
Read the guidesGitHub resources may require repository access.
LIMITED-TIME OFFER / CHALLENGE MDE
Tell us about a useful business application. AI interviews you to understand your business, then develops requirements, design and architecture for your review before building the final application. For selected challenges, we agree on scope and acceptance criteria, then work toward an application supported by test results and documented limitations.
Availability, timing and delivery terms must be confirmed before work begins; no closing date is published. Useful business applications only. No games, student projects, programming exercises, or toy applications. Submissions are selected based on business value, suitability, and reasonable scope.
See what AI believes about your business before that understanding becomes code.
BEGIN / GET STARTED
MDE gives AI coding agents such as Codex and Claude Code engineering knowledge, strategies, patterns and verification rules. Describe your business in natural language; AI interviews you, develops and maintains the specifications, and carries out the engineering work.
mde proceed and answer AI's questions. AI develops the specifications for your review.Node.js 24+ required. MDE is in active development and is not yet published to npm.
# After installing MDE from source
mde init my-app
cd my-app
mde workbenchmde proceed.