Onboard AI coding agents
like a software development team

Modular Chunks is the onboarding layer between you and your AI coding agent. You onboard it the way a CTO onboards a new development team: roles, scope, standards and guardrails up front — then the agent builds in small, checked steps, and you approve each one before the next begins.

The planning, delegation & verification platform

Modular Chunks is the deterministic layer on top of the agent, between you and whichever agent does the work.

User / subject matter expertDecides what gets built · approves every step
↓Decisions, research, sign-offPlain-language results to review↑
Who is this for?A · A group already working around itB · A broad group not fixing it
PlanningTurn an idea into a plan with intentional HITL questions.IdeaDiscoverBriefArchitecturePRDChunk graph
DelegationOne chunk at a time, through the tool you already use.CLIMCP serverChrome extensionManaged VMA developer
✓ Typed I/O✓ Guardrails✓ Tests● Your approval
VerificationEverything is checked against the eval checklist.Eval checklistHITL approval gateAuto-fixBlocking findingsAudit trail

The deterministic layer that sits on top of the AI coding agent

↓Typed, reviewed contractsCode, checked against the chunk↑
Onboards the AI coding agents you use
Claude CodeCursorWindsurfGitHub CopilotAiderOpenAI CodexGemini CLIClineRoo CodeAmazon Q DeveloperDevinReplit AgentLovableBoltv0Any MCP agent

The Agentic SDLC for the non-technical user

The Traditional SDLCA development team
01RequirementsPM / Analyst
02DesignDesigner
03CodeDevelopers · by hand
04TestQA Tester
05ReleaseDevOps
06MaintainSupport

Each arrow is a hand-off to a different person

A development team executes every stage, with AI helping as autocomplete.

The Agentic SDLC for the non-technical userA user or subject matter expert + agents
PlanBuildRun
01 · PlanDiscoverGuided questions → a brief
02 · PlanScopeArchitecture, PRD + UX/UI
03 · PlanChunksSplit into a typed chunk graph
04 · BuildDelegateCLI · MCP · Chrome Extension
05 · BuildAssembleHITL approval gate + auto-fix
06 · RunOperateFeature Builder · SDLC Tasks · Readiness

You approve at every step before the next one starts

The user or subject matter expert make decisions — that's where their knowledge and experience come in. Agents build inside typed, reviewed contracts.

User / SMEDecides
Modular ChunksOnboards & checks
AI Coding AgentBuilds

The thesis
Modular Determinism

Chunk 05
InstructionsWhat to build, step by step, in plain language
Typed I/OWhat goes in and what comes out, defined up front
Eval checklist4 checks the build has to pass
GuardrailsThe only files it is allowed to touch
Approved by you

Subscribe to the thesis

New essays and frameworks on Modular Determinism, plus product updates as Modular Chunks ships them.

Send me

No marketing, and unsubscribe any time.

You're on the list. Check your inbox to confirm.

For people who have the right to build and win — but don't know how to code

You bring what you know about the problem and make every call. Modular Chunks and your AI coding agent handle the build.

Non-technical users

Turn the idea in your head into working software, without writing code.

IdeaSoftware
Subject matter experts

Digitize your expertise and turn it into software you can sell.

ExpertiseProduct
Founders

Ship a deployable MVP, not a demo that drifts.

ScopeMVP
Product managers

Scope a feature and prototype it before it reaches engineering.

FeaturePrototype
Innovation teams & incubators

Prototype within your compliance, legal and marketing guidelines.

GuidelinesPrototype
Consulting firms

Customize the agentic SDLC for your clients’ experts.

Client expertsAgentic SDLC
Education institutions

Learn how to build software while shipping a real MVP.

LessonMVP
Technical teams

Ideate, write PRDs, prototype, and set the guidelines non-technical users build within.

PRDGuardrails

Build from where you already work

Drive Modular Chunks from your terminal, connect any agent over MCP, or ask Chunky beside any tab with the Chrome extension.

Knowledge transfer by design
The human's role is changing

SaaS era

Software was the service.

The human was the product — the operator.

People worked the tool: clicking, filling forms, moving data between systems.

→
Agent era

AI is the product.

The human is the supervisor.

People direct the work: setting requirements, judging results, signing off.

How your judgment carries into the next build

You supervise the build
01QuestionsPull decisions out of your head
02HITL gatesRecord each yes or no
03Process recommenderSuggests the next step
04GuidelinesYour standards
05PlaybooksReusable ways of working
Modular MemoryDesigned to carry your decisions into every new project, so the agent starts with the context it needs
Modular Memory
CorrectionsFeedback you gave on the agent’s work
DecisionsThe calls you made, and the reasons behind them
ReviewsWhat you approved, what you sent back, and why
AnswersYour replies to the HITL questions
ConventionsPatterns every new chunk follows

Roll it out across your team

For enterprises, consulting firms, incubators, accelerators and venture studios — give every team or cohort the same supervised way to build, by subscription, as a headless integration, or white-labeled as your own.

Bring your ideas
We'll onboard your agent

Modular Chunks is in closed beta. We review every request and onboard each user personally, so you build intentional software with AI coding agents from day one.

  1. 01Tell us the idea

    Request access with a line or two on what you want to build.

  2. 02We onboard you

    We review your request and set up your workspace with you.

  3. 03You start building

    Your AI coding agent builds; you approve every chunk.