“I’m not a real developer”
People with great ideas hold back because they fear their code isn’t “production-ready.” The idea dies before it starts.
Build for the joy of building
Ship your idea. Let AI help you write it. When it grows, ModernizeSpec gives any AI agent the context to refactor, restructure, and modernize — so you never have to start over.
npx modernizespec init 
The Real Problem
Too many ideas never get built
“I’ll rewrite it properly later”
Later never comes because there’s no map. The codebase grows, context scatters, and rewriting feels impossible.
“It’s too messy to touch”
AI agents encounter codebases with no structured understanding. Without context, even the best AI starts from zero every time.
Your Safety Net
Three layers. Zero gatekeeping.
Specification
When you’re ready to level up, drop JSON files in .agents/modernization/.
Bounded contexts, complexity scores, extraction plans. Any AI agent reads them instantly.
SDK + CLI
@modernizespec/sdk reads, validates, and generates spec files. The CLI wraps
it for terminal use. Zero lock-in — the spec works without the SDK.
Agent Bridge
Connect any AI coding agent to your codebase context: AGENTS.md (universal), Agent Skill (progressive), or MCP Server (deterministic tool access).
→ Integration GuideThe Difference
Build first. Structure later.
The old way
> "I have an idea, but I need to architect it properly first..."
[Spends 3 weeks on design documents]
[Debates framework choices]
[Sets up CI/CD before writing business logic]
[Writes tests for code that doesn't exist yet]
[Idea dies in planning. Nothing ships.]
Build fearlessly
> "I have an idea. Let me build it with AI."
✓ Ship a working prototype in days
✓ Validate the idea with real users
✓ Code is messy? That’s fine — it works
✓ When ready: modernizespec init
✓ AI agents read the spec, understand the codebase
✓ Refactor with confidence, not from scratch
How It Works
Build now. Modernize in five minutes.
1
Build your idea
Use any AI agent to turn your idea into working code. It doesn’t have to be pretty.
2
modernizespec init
When you’re ready, run the wizard. Answer questions about your codebase.
3
modernizespec analyze
The analyzer scans your code and maps bounded contexts, complexity hotspots, and dependencies.
4
Structured context appears
JSON files in .agents/modernization/ give any AI agent a map
of your codebase. No more starting from zero.
5
AI-assisted evolution
Agents understand your domains, complexity, and extraction order. Refactor with confidence — not from scratch.
Interactive Demo
Explore a real context map
This force-directed graph shows bounded contexts from the ERPNext case study. Node size reflects lines of code. Drag nodes to rearrange.
Proven on a Real System
Battle-tested on a 20-year-old codebase
316,000
Lines of Code
521
Entities
768
Endpoints
20
Years Old
7
Intern Teams
ERPNext is a 316K-line Python monolith — 20 years of organic growth, messy code, and real-world value. Seven intern teams used ModernizeSpec to map its bounded contexts, coupling scores, and complexity hotspots. Proof that any codebase can be understood and evolved.
→ Read the Case StudyPhilosophy
Bad code is good code
ரின் அட் கரை நல்லது
“Even reaching the shore poorly is still reaching the shore.” Bad code means someone tried. Someone built something. That matters more than clean architecture on day one.
Ideas over engineering
The hardest part is finding a problem worth solving. TDD, clean code, and “the right way” can wait. Ship the idea first. AI agents can help you build it — and later, help you rebuild it.
A safety net, not a gatekeeper
ModernizeSpec is the structured context that lets any AI agent understand your codebase when you’re ready to improve it. Not before. Not on anyone else’s timeline.
Roadmap
Making modernization effortless
Q1 2026
- File convention finalized
- Spec documentation site
- ERPNext reference implementation
- GitHub open-source launch
Q2 2026
- JSON Schemas published
- @modernizespec/core SDK
- First external adoption
Q3 2026
- Full SDK + CLI release
- MCP server for agent tooling
- Conference talks
Q4 2026
- Copilot integration support
- Cursor rules generation
- Consulting firm adoption