v1.0 · VS Code Marketplace TypeScript / JavaScript

Your codebase is drifting.
Now you can prove it.

ArchDrift is an architectural intelligence layer for TypeScript teams — detecting layer violations, cyclic dependencies, and technical debt with a financial cost model. The first tool to turn code health into a trackable drift metric.

AI generates code. ArchDrift makes sure it doesn't quietly break your architecture.

43
Real repos benchmarked
(React, Prisma, Supabase…)
5
Delivery surfaces
(VS Code, CLI, MCP, LSP…)
84%
TypeScript adoption
among developers (2025)
0
Direct competitors with
drift-over-time + MCP
archdrift scan . --full --output final_report.html
$ archdrift scan . --full --output final_report.html
◆ Building workspace dependency graph…
◆ Resolving TS path aliases + index barrels…
✓ Graph complete — 1,842 nodes · 4,391 edges

◆ Running Tarjan SCC cycle detection…
✗ 3 cyclic dependency clusters detected
└─ services/auth → utils/token → services/auth

◆ Scanning layer compliance rules…
⚠ 7 layer violations (domain → infra boundary)
◆ Detecting N+1 candidates in loops…
⚠ 12 N+1 candidates (mark as speculative)

◆ Drift Score
Previous scan ██████████░░░░░░░░░░ 4.8%
Current scan ███░░░░░░░░░░░░░░░░░░ 0.2% ↓ 4.6% improvement

◆ Cost attribution (git blame + CostConfig)…
✗ Estimated debt principal: $18,400
⚠ Monthly interest (delay cost): $920

✓ Report written → final_report.html
✓ Sections: Graph · Cost · Team Leaderboard · Lessons

Code quality tools measure
the wrong things.

Linters catch syntax. Type checkers catch types. Nothing tells you whether your architecture is eroding — until it already has.

🔥

Invisible Layer Violations

Domain code calling infrastructure directly. Service layer importing presentation logic. These violations compound silently across hundreds of commits, making large refactors progressively more expensive.

🔄

Cyclic Dependencies as Time Bombs

Circular imports lock teams into a ball of mud. They break tree-shaking, make testing impossible, and spread across files like mold — undetected until deployment breaks.

💸

Technical Debt Without a Price Tag

Engineering managers cannot prioritise refactors without financial framing. "We have tech debt" is ignored. "$18,400 principal accruing $920/month" triggers budget conversations.

📉

No Drift Measurement Over Time

Every existing tool gives a point-in-time snapshot. No tool tracks whether architectural quality is improving or degrading sprint-over-sprint — until ArchDrift.

One scan. Full architectural intelligence.

Core Metric

Drift Score™

The headline number no other tool tracks. A production-only percentage showing how far your architecture has moved from its baseline — improving or degrading over time.

0.0
% DRIFT
↓ from 4.8% last scan
Graph Engine

Dependency Graph + SCC

Full workspace graph with TS path alias resolution, barrel handling, and Tarjan SCC cycle detection. Cyclic clusters injected as first-class issues.

Zero-Trust

Skeleton Analyzer

Analyse architecture without source code. Hash-based skeleton files <10% of source size. Orphan detection with entry point exclusions. Unique competitive differentiator.

skeleton/auth.json 2.4 KB
source/auth.ts 28 KB
compression ratio 8.6%
Financial Layer

Debt Cost Attribution

The first architectural tool that speaks CFO language. Git blame maps violations to teams. CostConfig converts findings to dollar figures — principal debt + monthly interest accrual.

Layer violations × hourly rate $8,200
Cyclic dependency remediation $6,400
N+1 candidate review cost $3,800
Total debt principal $18,400
Monthly interest (delay cost) $920 / mo
AI-Era

MCP Interceptor

Blocks cyclic/layer violations before AI agents commit them. First-mover in agentic architecture guardrails.

BLOCK · cyclic dependency
BLOCK · layer violation
WARN · N+1 candidate
PASS · valid import
Editor

VS Code Extension

Inline diagnostics + "Generate Workspace Report" command. Published on VS Code Marketplace. Benchmarked on 43 real repos.

React Prisma Supabase Angular
CLI + CI/CD

CLI Pipeline

Integrate into any CI pipeline. archdrift scan . --full outputs HTML + JSON. Windows-compatible with archdrift.cmd helper.

archdrift scan .
archdrift skeleton ./src
archdrift diff HEAD~10

5 production phases. All shipped.

Every phase was built test-first with integration verification. No mock data, no stubs in production paths.

Phase 01

Core Graph Engine

  • Bidirectional edge tracking
  • TS path alias + barrel resolution
  • Tarjan SCC cycle detection
  • Cyclic issues injected into analysis
Tests passed
Phase 02

Zero-Trust Skeleton

  • Hash + metric extraction (no source)
  • VirtualAnalyzer from skeleton data
  • Orphan detection with exclusions
  • Skeleton <10% of source size
Tests passed
Phase 03

AI Immune System (MCP)

  • MCP server with check_architecture
  • BLOCK on cycles + layer violations
  • WARN on N+1 candidates
  • Interceptor test suite
Tests passed
Phase 04

Financial + Governance

  • CostConfig + calculateDebtCost
  • Lazy git blame line attribution
  • Team mapping via .archdriftrc.json
  • Team leaderboard generation
Tests passed
Phase 05

Visualization + Trends

  • HTML report (Graph, Cost, Leaderboard)
  • DiffEngine with driftVelocity
  • CLI scan with --full HTML output
  • archdrift.cmd Windows helper
Final report verified

A gap no existing tool fills.

84%
TypeScript adoption among professional developers
// Stack Overflow Developer Survey 2025
$85B+
Global developer tooling market by 2027
// Developer tools segment, VC-backed 2024–2027
$0
Direct competitors with drift tracking + MCP agentic guardrails
// Competitive analysis: ArchLintor, Architex, SonarQube

Competitive landscape

Tool Layer Cycles Drift Cost MCP
ArchDrift
ArchLintor
Architex ~
SonarQube ~ ~
ESLint

✓ native ~ partial/paid ✗ absent

What makes this hard to copy.

📐

Engine/Extension Separation

archdrift-engine has zero VS Code API dependencies. The same core powers CLI, MCP, LSP, and future integrations — a platform, not just a plugin.

⏱️

Drift-as-Metric Pioneer

The concept of architectural drift as a trackable, comparable number is novel. First-mover advantage compounds as teams build institutional workflows around the score.

🤖

MCP Guardrail First-Mover

With AI agents writing more code daily, architectural guardrails at the MCP layer become critical infrastructure. ArchDrift is the only tool positioned here today.

🔬

43-Repo Benchmark Corpus

Calibrated against React, Prisma, Supabase, Storybook, Angular — not toy examples. This benchmark corpus is proprietary signal that trains scoring credibility.

🤖

Built for the AI Coding Era

AI writes code with no architectural memory. Every Cursor or Copilot session drifts slightly further from your intended structure — silently. ArchDrift is the watchdog that AI can't be for itself. As AI coding adoption compounds, so does the drift problem — and so does ArchDrift's relevance.

50 real repositories. Scanned cold.

ArchDrift ran against 50 production-grade open-source repos — axios, React, Node.js, TypeScript, VS Code, Bun, NestJS, and more — with zero configuration. These are the raw results.

50
Repos scanned
3.3M+
Lines of code analyzed
1,009+
Violations detected
6
Perfect 100% SII
92%
Average integrity score
0
Config files required
Structural Integrity Index — 50 repos scanned
100% 95% 90% 85% 83% axios: 100.0% express: 100.0% fastify: 100.0% lucide: 100.0% deno: 100.0% solid: 100.0% trpc: 96.9% shadcn-ui: 99.0% create-t3-app: 98.7% pnpm: 98.7% svelte: 97.6% angular: 97.7% hono: 96.6% nest: 95.9% zod: 95.8% tanstack-query: 93.9% react: 93.7% node: 93.2% remix: 92.1% vscode: 91.6% vite: 91.0% playwright: 89.2% bun: 83.3% typescript: 75.6% esbuild: 75.4% lodash: 74.0% nocodb: 73.7% directus: 72.0% axios express fastify lucide deno solid trpc shadcn-ui create-t3-app pnpm svelte angular hono nest zod tanstack-query react node remix vscode vite playwright bun typescript esbuild lodash nocodb directus
100% — Perfect
95–99% — Strong
85–94% — Moderate drift
75–84% — High drift
<75% — Critical
Repository SII Score Violations Prod LOC N+1 God Classes Layer Violations
axios 100.0% 0 661
express 100.0% 0 1,139
fastify 100.0% 0
lucide 100.0% 0 3,450
deno 100.0% 0
solid 100.0% 1 5,110 1
trpc 96.9% 2 26,887 1 1
shadcn-ui 99.0% 1 23,682 1
create-t3-app 98.7% 2 15,146 2
pnpm 98.7% 9 95,837 3 6
svelte 97.6% 6 45,557 6
angular 97.7% 36 255,652 4 31 1
hono 96.6% 7 37,570 2 5
nest 95.9% 7 40,585 2 5
zod 95.8% 12 48,985 12
tanstack-query 93.9% 14 42,777 10 4
react 93.7% 51 253,532 2 49
node 93.2% 41 145,609 2 39
remix 92.1% 5 23,814 5
vscode 91.6% 270 1,037,500 40 230
vite 91.0% 14 47,054 1 13
playwright 89.2% 29 215,035 8 21
bun 83.3% 39 110,780 6 33
typescript 75.6% 68 258,489 68
esbuild 75.4% 1 5,000 1
lodash 74.0% 1 80,000 1
nocodb 73.7% 213 256,827 157 56
directus 72.0% 180 180,000 15 40 125
Most extreme finding
TypeScript's checker.ts: 44,054 lines
A single file in the TypeScript compiler is 44K lines — ArchDrift flagged it as a critical monolith. The overall TypeScript repo scores 75.6% SII across 258K LOC with 68 god classes.
Architecture champion
axios, express, lucide — perfect clean
Six repos scored 100% with zero weighted violations. These are the reference architectures — lean, modular, zero drift. ArchDrift can identify and benchmark against them.
Most violations
VS Code: 270 violations in 1M LOC
VS Code's 1,037,500 production lines contain 230 god classes and 40 N+1 query patterns — scoring 91.6% SII. The drift is manageable but measurable and growing.
Layer drift hotspot
Directus: 125 layer violations
Directus has 125 service-to-api layer violations — the most in the benchmark. Layer violations are the hardest drift to untangle; they break dependency inversion and lock teams into monolith patterns.

Let's talk about the architecture of growth.

ArchDrift is raising a pre-seed round to accelerate distribution, CLI-first monetisation, and enterprise team features. Request the full deck or book a 30-minute demo.

Direct: prajeesh@meethale.com