How Valarchi Works

Six steps. One loop. Here's how it all fits together.

Milestone 1

Analyze

PRD

App Structure

Wireframes

Tech Plan

Code

↩ re-analysis picks up new code → Milestone 2 begins

Step 1

It starts with your codebase

Start by connecting your codebase through Valarchi's GitHub app. Valarchi extracts a semantic snapshot of your product that all downstream AI agents use as their foundation.


Step 2

PRD — scope the milestone

Describe what you want to build for this milestone in a PRD template you like. Use Valarchi AI to generate or refine sections. Discuss tradeoffs, stress-test scope, or explore edge cases with AI powered chat that understands your existing product.


Step 3

App Structure — what changes

Use this AI agent to identify what screens and user flows need to be updated based on the PRD.


Step 4

Wireframes — hi-fi, token-aware

Use this AI agent to generate hi-fi interactive wireframes for each screen identified in the app structure step.


Step 5

Tech Plan — architecture + task specs

Use this AI agent to generate an engineering architecture overview and a task breakdown for the milestone.


Step 6

Coding agents pick it up via MCP

Connect your AI coding agent to Valarchi's MCP server to get each individual task specification and execute the task.


Then the loop repeats

After completing a milestone, kick off a re-analysis to update the semantic snapshot of your product with the new changes. The upcoming milestones will use this new snapshot as a baseline to create planning artifacts. The iteration loop continues.