ITERATIVEIMPACT
MVPFOUNDRY

A structured workflow for turning client briefs into deployed products.

MVP Foundry is a personal toolchain combining Claude AI, an autonomous coding agent, and a repeatable eight-step process - from requirements capture to running code on Vercel.

Built by Iterative Impact. Not a product for sale - a methodology in active use.

How it works

Eight steps from client brief to deployed product. Expand any step to see exactly what it produces.

1

Client Brief

Upload or describe the project. AI extracts structured requirements, identifies user types, key features, constraints, and success metrics. This document is the source of truth for every step that follows.

2

Architecture Options

AI generates 3-4 genuinely different architecture approaches - not minor variations. Each gets a PlantUML C4 diagram enriched with any detected patterns (SaaS, data stack, web intelligence). A Q&A panel captures decisions before committing.

3

Pre-Sales Pack

optional

Client-facing documents generated from brief and architecture options. Architecture overview with C4 diagram, a full proposal document, and Canva content blocks for a pitch deck. Shareable before the client commits to build.

4

Detailed Architecture

Five standard diagrams at the right level of abstraction: container diagram (C4 L2), business process flows, deployment architecture, entity relationship overview, and key data flows. An AI critique panel automatically reviews the output against the brief - flagging scope drift, missing diagrams, and BDUF violations before you share it.

5

Select Components

optional

Choose which components the AI Build Agent should implement. Deselect infrastructure-only items, third-party config, or anything better done manually. Skip this step entirely to have the agent build everything.

6

PRD & Repository

AI converts the architecture into atomic, dependency-ordered GitHub issues - one per feature, nothing blocked. Creates the repository with full scaffolding, CI/CD pipeline, autonomous agent workflow, and project memory file (agents.md).

7

AI Build Agent

An autonomous coding agent running in GitHub Actions works through issues independently. It generates code against the architecture spec, runs tests, self-corrects on failure, creates pull requests, auto-merges, and chains to the next issue without human input.

8

Review & Deploy

Monitor progress via GitHub, review merged pull requests, and deploy to production when a milestone is ready. Vercel auto-deploys on merge - no manual deployment steps.