StackSketch

Turn any codebase into a beautiful architecture map in one command.

StackSketch is a zero-dependency CLI that scans a repository locally, detects languages/frameworks, extracts best-effort imports, builds a dependency graph for supported ecosystems, and exports a polished HTML report, Markdown summary, or JSON dataset.

It is designed for the moment developers love to share: new project onboarding, PR context, architecture reviews, open-source READMEs, and “look how clean this codebase is” screenshots.

Why it exists

Most code visualization tools are either:

• tied to one language,
• locked behind a SaaS dashboard,
• heavy to install,
• or too abstract to be useful.

StackSketch takes the opposite approach:

• runs locally,
• uploads nothing,
• has no runtime dependencies,
• works across JavaScript, TypeScript, Python, Go, Rust, Java, C#, PHP, Ruby, and more for language detection, with local dependency edges for JavaScript/TypeScript, Dart, Python, Rust, Go, and CSS assets,
• produces a visual report that is useful and shareable.

Quick start

npx stacksketch-cli .

This creates stacksketch.html in the current directory. Open it in a browser to explore the interactive architecture map.

npx stacksketch-cli ./my-app --open

Install globally

npm install -g stacksketch-cli
stacksketch .

Output formats

HTML report

stacksketch . --open

Generates an interactive, self-contained HTML report with:

• language distribution,
• framework detection,
• import graph,
• top files,
• searchable nodes,
• local and external dependency edges,
• SVG download,
• Markdown copy button.

Markdown report

stacksketch . --format md --output ARCHITECTURE.md

Useful for PRs, RFCs, onboarding docs, and README sections.

JSON dataset

stacksketch . --format json --output stacksketch.json

Useful for CI, dashboards, custom renderers, or AI agent context.

CLI reference

stacksketch [root] [options]

Option	Description
-o, --output <path>	Output file. Defaults to stacksketch.html.
--format <html|md|json>	Output format. Defaults to file extension or HTML.
--html	Write stacksketch.html.
--json	Write stacksketch.json.
--md, --markdown	Write stacksketch.md.
--open	Open the HTML report after generation.
--title <title>	Report title.
--max-files <number>	Max source files to scan. Default: 500.
--ignore <pattern>	Extra ignore pattern. Repeatable.
--include <pattern>	Extra include pattern. Repeatable.
-h, --help	Show help.
-v, --version	Show version.

Examples

Scan the current repo:

stacksketch .

Scan a specific repo and open the report:

stacksketch ./packages/web --open

Generate a Markdown architecture summary:

stacksketch . --format md --output docs/architecture.md

Limit the scan to a large monorepo:

stacksketch . --max-files 1500

Ignore generated folders:

stacksketch . --ignore generated --ignore dist

Include only source folders:

stacksketch . --include src --include packages

What StackSketch detects

Languages

JavaScript, TypeScript, Dart, Vue, Svelte, Python, Go, Rust, Java, C#, PHP, Ruby, Swift, Kotlin, Scala, C, C++, CSS, SCSS, HTML, JSON, YAML, TOML, Markdown, SQL, Shell, PowerShell, Dockerfile, Terraform, and Lua.

Frameworks

React, Next.js, Vite, Nuxt, Svelte, Vue, Flutter, Express, NestJS, Fastify, Django, Flask, FastAPI, Tailwind CSS, Actix Web, Rocket, Gin, Echo, and Spring Boot.

Graph signals

• local imports for supported languages,
• external dependency signals for supported import styles,
• top files by LOC/import/export weight,
• language distribution,
• directory structure,
• file size and line count,
• best-effort exported symbols,

What makes it different

Local-first

StackSketch reads files on disk and writes a static report. It does not require an API key, cloud account, browser extension, or repository upload.

Zero runtime dependencies

The CLI is intentionally dependency-free. That makes installation fast, packaging simple, and CI usage reliable.

Built for screenshots

The HTML report is designed to look good immediately. It is the kind of output people can drop into a README, PR, landing page, or social post without extra design work.

Useful for AI workflows

The JSON output is a compact map of a repository that can be used as context for AI agents, codebase summarizers, onboarding bots, and architecture review pipelines.

Development

npm install --ignore-scripts
npm run check
npm test
npm run smoke
npm start

npm start scans the StackSketch repository itself and writes stacksketch.html.

Before publishing, run:

npm run check
npm test
npm run smoke
npm pack --dry-run
DRY_RUN=1 ./scripts/publish.sh

Packaging

npm pack
npm publish

The package includes the CLI source, scripts, README, changelog, contributing guide, and license.

Current status

StackSketch is usable as a local CLI and npm package. The project is not yet a full multi-language AST parser: language detection is broad, while local dependency resolution is strongest for JavaScript/TypeScript, Dart, Python, Rust, Go, and CSS assets.

Generated HTML and Markdown reports use display-friendly project names instead of embedding absolute local filesystem paths.

Roadmap

Planned work:

• GitHub URL input.
• Graphviz export.
• Plugin system for custom parsers.
• CI badge generation.
• Multi-repo monorepo views.
• Architecture smell detection.
• Local LLM-assisted summary mode with explicit opt-in.

Contributing

See CONTRIBUTING.md for development commands, test expectations, and contribution areas.

License

MIT