aichitect

---

AI tools are all over the place. Every week there's a new framework, a new model, a new "essential" addition to your stack. AIchitect gives you a structured, visual map of the ecosystem — 256 tools across 21 categories, with 35 curated stacks and a live signal feed — so you can pick the right stack based on data, not hype.

Features

Explore — the full ecosystem as an interactive graph

Browse all 256 tools as a force-directed graph. Filter by category or relationship type, search by name, and switch between three view modes:

• 2D Grid — clean, scannable card layout
• 2D Layers — swimlane view organized by stack layer (Development → AI Logic → Models & Infra → Tooling)
• 3D — rotatable Three.js force graph with orbit controls

Stacks — 35 mission-driven starting points

Stacks lead with the specific situation they solve, list what was explicitly rejected and why, and tell you exactly when to move on. Organized across 5 decision clusters:

Cluster	Tagline	Stacks
Build	Ship this week	9
Automate	AI does the work	8
Ship & Harden	Make it trustworthy	6
Understand	Data is the product	5
Comply & Restrict	Nothing leaves the building	5

Each stack includes the situation, why each tool was picked, what was rejected (and why), kill conditions, and the next stack to graduate to.

Builder — design your own stack

Pick one tool per slot and watch your stack wire together with live integration edges. Share your exact stack via a single URL (?s=cursor,langgraph,openai-api,…).

Simulate — project cost, latency, and breaking points

Drag a few sliders and watch the chart move: per-token cost with prompt-caching and batch-API discounts, token-aware latency, a shadow-stack A/B comparison, and a provider ranking at your specific workload. Built on a public per-tool cost-model catalog.

Compare — side-by-side tool analysis

Compare any two tools head-to-head: category, pricing, OSS vs. SaaS, GitHub stars, integrations, and a plain-language summary of the tradeoffs. Shareable via URL (/compare/cursor/claude-code).

Genome — analyse your existing stack

Paste your dependency files (package.json, requirements.txt, .env, etc.) and Genome detects which AI tools you're already using, scores your stack's fitness, surfaces gaps, and offers an adversarial challenge or AI roast of every tool choice.

Pulse — 30-day ecosystem momentum

A category-level heatmap of which parts of the AI tools landscape are heating up vs. cooling down — built from the nightly health-sync, benchmark, and incident pipelines.

Activity — what changed today

Live, filterable feed of tool events: health-score shifts, star milestones, stale transitions, archive detections, pricing changes, benchmark drift, and incidents — fed by the cron jobs that watch every tracked vendor.

Find My Stack — short quiz → recommended stack

Four questions (role, use case, budget, priorities) → a curated stack recommendation that maps into Builder and Compare.

Make a Case — defend your stack

Generates a structured argument for a given stack you can share with a teammate or stakeholder, including rejected alternatives and exit conditions.

An experimental MCP server endpoint (/api/mcp) is also exposed at https://aichitect.dev/api/mcp — see the setup guide. Not actively developed.

Tech Stack

Layer	Technology
Framework	Next.js 16 (App Router, Turbopack)
Language	TypeScript · React 19
2D graph	React Flow v11 + dagre + ELK
3D graph	react-force-graph-3d + Three.js
Styling	Tailwind CSS v4
Data layer	Supabase Postgres (auth + tool events) · static JSON
AI features	Google AI (@google/generative-ai) for Roast and Genome Challenge
Activity	Cloudflare cron triggers → health sync (nightly), benchmarks (weekly), incidents (hourly)
Dev runtime	Docker + Docker Compose

Getting Started

All dev runs through Docker — no local Node.js required.

# Clone
git clone https://github.com/letsgogeeky/aichitect.git
cd aichitect

# Start dev server (applies local migrations, seeds DB, tails logs)
make run

# Stop
make down

Requires Docker and Docker Compose. The dev server is exposed on http://localhost:3002 (mapped from container 3000).

Common commands

make run              # Start dev server (foreground, hot reload)
make down             # Stop and remove containers
make rebuild          # Full rebuild after adding new packages
make logs             # Tail container logs
make shell            # Open shell inside the running container
make typecheck        # Run tsc --noEmit
make lint             # Run ESLint
make test             # Run Vitest suite
make check            # lint + typecheck + test in one shot
make sync-counts      # Sync tool/category/stack counts in README + CLAUDE.md
make seed-validate    # Validate JSON data integrity without a DB
make db-push-local    # Apply pending migrations to local Postgres
make db-push          # Promote verified migrations to remote Supabase
make db-migrate name=X  # Scaffold a new migration file

Project structure

app/
  page.tsx           landing page (hero, view cards, live signals, footer)
  explore/           graph view (2D grid / swimlanes / 3D)
  stacks/            33 mission-driven stacks across 5 clusters
  builder/           pick-per-slot stack builder
  simulate/          cost / latency / breaking-points simulator
  compare/[a]/[b]/   side-by-side tool comparison
  genome/            paste-your-deps stack-fitness scorer
  pulse/             30-day category momentum heatmap
  feed/              live tool-event activity feed
  case/              "make a case for your stack" generator
  match/             4-question stack quiz
  mcp/               MCP setup guide
  tool/[toolId]/     per-tool detail pages
  category/[id]/     per-category landing pages
  profile/[user]/    public user profiles (GitHub auth)
  api/               cron, MCP, roast/challenge endpoints
  badge/             shields.io-style SVG badge endpoint
data/
  tools.json         256 tools
  relationships.json 553 edges (integrates-with / commonly-paired-with / competes-with)
  stacks.json        35 curated stacks with flow edges
  slots.json         22 builder slot types
lib/
  types.ts           single source of truth for all interfaces + category colors
  simulate.ts        cost / latency simulation core
  genomeAnalysis.ts  dependency-file → stack genome detection
  health.ts          GitHub-signal → health score
  pulse.ts           category-momentum aggregation
  data-loaders.ts    typed JSON loaders
components/
  graph/             ExploreGraph (2D/3D), ToolNode, LaneLabel, EnrichedEdge
  panels/            FilterPanel, DetailPanel, ComparisonPanel, StackHealthPanel
  ui/                Navbar, Logo, MyStackTray, modals, walkthrough
hooks/
  useBuilderState · useComparisonMode · useUser · useIsMobile
scripts/
  sync-counts.mjs    patches hardcoded counts in README + CLAUDE.md
  seed-db.ts         seed remote Supabase (idempotent upsert)
  seed-db-local.ts   seed local Docker Postgres
supabase/migrations/ DB schema history

See CLAUDE.md for the architectural deep-dive (edge-runtime rules, data integrity constraints, type-system contracts).

Contributing

Contributions are welcome — tools, stacks, bug fixes, and new features.

Adding a tool

1. Add an entry to data/tools.json:

{
  "id": "my-tool",
  "name": "My Tool",
  "category": "agent-frameworks",
  "tagline": "One sentence description",
  "description": "Longer description shown in the detail panel.",
  "type": "oss",
  "pricing": { "free_tier": true, "plans": [] },
  "github_stars": 12000,
  "slot": "orchestration",
  "prominent": false,
  "website_url": "https://mytool.dev",
  "github_url": "https://github.com/org/repo"
}

2. Optionally add edges in data/relationships.json:

{ "source": "my-tool", "target": "langchain", "type": "integrates-with" }

3. Run make seed-validate to catch broken references; make sync-counts to refresh the badges.

Adding a stack

Add an entry to data/stacks.json. The full schema is in lib/types.ts (Stack). cluster must be one of: build · automate · ship · comply · understand. All tool IDs in tools, not_in_stack, flow.from, and flow.to must exist in data/tools.json.

Opening an issue

Found a missing tool, broken edge, or UI bug? Open an issue — all feedback is welcome.

Pull requests

1. Fork the repo
2. Create a branch: git checkout -b feat/my-change
3. Make your changes and verify with make check (lint + typecheck + tests)
4. Open a PR with a clear description of what changed and why

License

MIT — free to use, modify, and distribute.

---