Generate Tech Stack

io.github.askuma由 GitHub 帳號 askuma 發布v0.2.2更新於 Oct 24, 2026

Visual tech-stack inventory of any codebase: languages, frameworks, databases, AI SDKs, infra.

發布方自己的服務：Tech.eu Funding Explorer

已驗證STDIO僅桌面開發者工具資料與分析

服務詳細資訊

識別碼: io.github.askuma/generate-tech-stack
傳輸: STDIO
執行環境: Desktop only
信任: Verified
發布於: Sep 16, 2026

安全提示

MCP 服務會接收對話中的工具呼叫請求，並可能執行外部操作。啟用前請先了解該服務能做什麼，並只授予它所需的權限。

註冊中心已驗證此條目所屬的 GitHub 帳號 askuma。這並不表明該帳號就是產品方。

SourceWeft 的檢查只涵蓋條目是否完整、端點是否回應；不審查伺服器的程式碼，也不審查其工具的行為。

在 README 中新增徽章

向讀者展示該 MCP 服務已收錄於 SourceWeft，徽章連結到本頁。

[![Listed on SourceWeft](https://sourceweft.com/badge/mcp/io.github.askuma%2Fgenerate-tech-stack.svg)](https://sourceweft.com/mcp/io.github.askuma%2Fgenerate-tech-stack?utm_source=readme&utm_medium=badge)

概覽

AI 產生的概覽

掃描本機程式碼庫，產生呈現語言、框架、資料庫、AI SDK 與基礎架構的視覺化 TECH_STACK.html 報告。

可以做什麼: 提供兩個 MCP 工具：generate_tech_stack 會掃描專案、寫出自包含的 TECH_STACK.html 報告並在瀏覽器開啟；list_tech_stack 回傳 JSON 摘要且不寫入檔案。偵測範圍涵蓋 requirements.txt、pyproject.toml、package.json、go.mod、Cargo.toml、Dockerfile、docker-compose.yml 及 CI 工作流程等相依與設定檔，並將結果歸入語言、Web 框架、資料庫、AI SDK、可觀測性、測試、安全、基礎架構等類別。報告包含統計列、架構圖、長條圖、工具卡片、徽章圖例與導覽導引。
適用情境: 適合需要快速取得專案技術棧視覺化清單的情境，例如新人上手、撰寫文件或程式碼審查，也適合讓助理不必逐一人工閱讀清單檔就能總結倉庫相依與工具鏈。
執行需求: 透過 stdio 在本機執行。需安裝 PyPI 套件 generate-tech-stack-mcp（若直接執行伺服器腳本則需 mcp 套件），要求 Python 3.8 以上；掃描腳本本身沒有第三方相依。未宣告任何帳號、API 金鑰或環境變數。

安裝前請注意

generate_tech_stack 會在掃描的專案目錄中寫入 TECH_STACK.html 並在瀏覽器開啟，請只對允許變更的目錄使用。產生的頁面會從 Google Fonts 載入字型，開啟時會連線至第三方；除此之外報告為自包含檔案。

安裝

在 SourceWeft 中

開啟儀表板中的 Generate Tech Stack，將其新增到工作區。
為需要使用其工具的對話啟用該服務。

Desktop only，透過 STDIO。 STDIO 服務會啟動本機處理程序，因此需要 SourceWeft 桌面主機。

其他 MCP 客戶端

參照儲存庫中的啟動說明。

README

generate-tech-stack

Scan any project and generate a visual TECH_STACK.html page — light/dark theme, auto-adapting, zero config.

Works as a Claude Code skill, MCP server, or GitHub Copilot Extension.

Live demo → — generated from fastapi/full-stack-fastapi-template, unmodified.

[Running generate-tech-stack against fastapi/full-stack-fastapi-template, from pip install to the generated page]

(real CLI output, unscripted — static screenshot if you'd rather not autoplay)

What it produces

Every generated page contains:

Section	Description
Stat row	Total tools · Categories · AI Backends · Data Stores
Architecture diagram	Layered flow diagram (Consumer → API → AI/NLP → Data/Obs/Frontend)
Bar chart	Horizontal bars per category, colour-matched
Tool cards	One card per category; each tool shows a dot, name, description, and badge
Badge legend	Explains `pip`, `dep`, `optional`, `core`, `deploy`, `ci`, etc.
Footer	Project name · tool count · generation date
Guided tour	Spotlight walkthrough of every section, shown automatically the first time a report is opened; replay anytime with the `?` button
Theme toggle	Sun/moon button next to the tour button switches between dark and light mode

Repository layout

~/.claude/skills/generate-tech-stack/├── SKILL.md                  ← Claude Code skill definition├── INSTALL.md                ← detailed per-platform installation guide├── README.md                 ← this file├── scripts/│   └── analyze.py            ← core scanner + HTML renderer (no dependencies)├── mcp/│   ├── server.py             ← MCP stdio server (pip install mcp)│   └── requirements.txt└── copilot/    ├── index.js              ← GitHub Copilot Extension (Express)    ├── package.json    └── openai_function.json  ← OpenAI / Antigravity function definition

~/.claude/skills/generate-tech-stack/├── SKILL.md                  ← Claude Code skill definition├── INSTALL.md                ← detailed per-platform installation guide├── README.md                 ← this file├── scripts/│   └── analyze.py            ← core scanner + HTML renderer (no dependencies)├── mcp/│   ├── server.py             ← MCP stdio server (pip install mcp)│   └── requirements.txt└── copilot/    ├── index.js              ← GitHub Copilot Extension (Express)    ├── package.json    └── openai_function.json  ← OpenAI / Antigravity function definition

Usage

pip (CLI + MCP server)

bash

pip install generate-tech-stack-mcp
generate-tech-stack . TECH_STACK.html    # CLI: scan and write the reportgenerate-tech-stack-mcp                   # stdio MCP server

pip install generate-tech-stack-mcp
generate-tech-stack . TECH_STACK.html    # CLI: scan and write the reportgenerate-tech-stack-mcp                   # stdio MCP server

With pip installed, any MCP host config reduces to:

Claude Code

Run it from any project directory. The skill calls scripts/analyze.py and opens the result in your browser.

MCP (Claude Desktop, VS Code, Cursor, Zed, Windsurf, Continue)

Add to your host's MCP config (replace the path with your actual home directory):

json

{  "mcpServers": {    "generate-tech-stack": {      "command": "python3",      "args": ["/home/<you>/.claude/skills/generate-tech-stack/mcp/server.py"]    }  }}

{  "mcpServers": {    "generate-tech-stack": {      "command": "python3",      "args": ["/home/<you>/.claude/skills/generate-tech-stack/mcp/server.py"]    }  }}

Then ask: generate my tech stack or /generate-tech-stack.

MCP tools exposed:

generate_tech_stack — scans a project, writes TECH_STACK.html, opens in browser
list_tech_stack — returns a JSON summary, no file written

GitHub Copilot Extension

bash

cd copilotnpm installnpm start        # listens on port 3000ngrok http 3000  # expose for GitHub to reach

cd copilotnpm installnpm start        # listens on port 3000ngrok http 3000  # expose for GitHub to reach

Register a GitHub App with Copilot Extension enabled, set the Agent URL to https://your-url/agent, and install it on your account. Then in Copilot Chat:

Command line (standalone)

bash

python3 ~/.claude/skills/generate-tech-stack/scripts/analyze.py /path/to/project# output: /path/to/project/TECH_STACK.html
# custom output path:python3 scripts/analyze.py . ~/Desktop/TECH_STACK.html

python3 ~/.claude/skills/generate-tech-stack/scripts/analyze.py /path/to/project# output: /path/to/project/TECH_STACK.html
# custom output path:python3 scripts/analyze.py . ~/Desktop/TECH_STACK.html

analyze.py has no third-party dependencies — just Python 3.8+.

What gets detected

Source file	Detected tools
`requirements.txt` / `pyproject.toml`	Python packages (web, DB, AI, testing, observability, security…)
`package.json`	Node / npm packages (frameworks, frontend, DB drivers, tooling)
`go.mod`	Go language
`Cargo.toml`	Rust language
`pom.xml` / `build.gradle`	Java / Kotlin
`Gemfile`	Ruby
`composer.json`	PHP
`*/.py` source	Optional/dynamic SDKs via `importlib.find_spec()` patterns
`docker-compose.yml` / `.env`	PostgreSQL, Redis, MongoDB, SQLite connection strings
`Dockerfile`	Docker
`docker-compose.yml`	Docker Compose
`.github/workflows/`	GitHub Actions
`.gitlab-ci.yml`	GitLab CI
`alembic.ini`	Alembic migrations
`nginx.conf` / `Caddyfile`	Reverse proxy
`tsconfig.json` / `src/*/.ts`	TypeScript

Detected categories

Category	Colour	Examples
Language & Runtime	Green	Python, Go, Rust, TypeScript
Web / API Framework	Purple	FastAPI, Express, Django, Next.js
Database / Storage	Green	SQLAlchemy, Prisma, Redis, ChromaDB
AI SDKs	Blue	OpenAI, Anthropic, LangChain, GuardrailsAI, NVIDIA NeMo, Presidio
NLP / ML	Teal	spaCy, Transformers, Sentence Transformers
Observability	Teal	Prometheus, OpenTelemetry, Sentry, Loguru
Testing	Yellow	pytest, Jest, Cypress, Playwright
Security / Auth	Rose	PyJWT, bcrypt, Authlib, Helmet
Infrastructure / Deploy	Orange	Docker, Kubernetes, Celery, Boto3
Frontend / Dashboard	Gray	React, Vue, Tailwind, Recharts
Messaging / Comms	Blue	Kafka, RabbitMQ, Socket.io
Dev Tools	Gray	ESLint, Prettier, Vite, TypeScript

Design

Dark by default with a light-mode toggle (top-right corner, no persistence — resets to dark on reload). Fonts: IBM Plex Sans (body) + JetBrains Mono (code/badges), loaded from Google Fonts. Vanilla JS powers the guided tour and theme toggle only — everything else is plain HTML + CSS. Self-contained single file, opens in any browser offline.