Skip to main content
Glama

project-scaffold

An MCP server that scaffolds full-stack projects with consistent structure, Docker setup, CI/CD pipelines, and database configuration. Built with FastMCP.

Alpha — This project is under active development. APIs and generated output may change between versions. Contributions are welcome! See CONTRIBUTING.md.

What it generates

Every scaffolded project comes with:

  • FastAPI backend with async SQLAlchemy and Pydantic v2

  • PostgreSQL database with Docker Compose (includes pgAdmin)

  • Frontend — React (Vite + TypeScript) or HTMX (Tailwind CSS)

  • Alembic database migrations (optional)

  • JWT authentication boilerplate (optional)

  • CI/CD — GitHub Actions for Azure, or Render.com blueprint

  • Makefile with common dev commands

Related MCP server: Builder-Proj-MCP Server

Template combinations

Project type

Frontend

Deploy target

Repo structure

work

react

Azure (App Service + Static Web Apps)

Multi-repo (API + frontend)

work

htmx

Azure (App Service)

Single repo

personal

react

Render.com

Multi-repo (API + frontend)

personal

htmx

Render.com

Single repo

Requirements

  • Python 3.12+

  • uv (recommended) or pip

Installation

From GitHub (pip)

pip install git+https://github.com/dawiegriesel/mcp-dev.git

From GitHub (uv)

uv pip install git+https://github.com/dawiegriesel/mcp-dev.git

Local development

git clone https://github.com/dawiegriesel/mcp-dev.git
cd mcp-dev
uv sync

For development dependencies (pytest, ruff):

uv sync --extra dev

Installing the MCP server

Claude Code

claude mcp add project-scaffold -- uvx --from "git+https://github.com/dawiegriesel/mcp-dev.git" project-scaffold

Or if installed locally:

claude mcp add project-scaffold -- uv run --directory /path/to/mcp-dev project-scaffold

Project-scoped config (local development)

For local development, create a .mcp.json in the repo root to register the server automatically when you open the repo in Claude Code — no manual claude mcp add needed:

{
  "mcpServers": {
    "project-scaffold": {
      "command": "uv",
      "args": ["run", "--directory", "/your/local/path/mcp-dev", "project-scaffold"]
    }
  }
}

Replace the --directory value with your actual clone path. This file is gitignored to avoid committing local paths.

Claude Desktop

Add this to your Claude Desktop config file (~/Library/Application Support/Claude/claude_desktop_config.json on macOS, %APPDATA%\Claude\claude_desktop_config.json on Windows):

{
  "mcpServers": {
    "project-scaffold": {
      "command": "uvx",
      "args": [
        "--from", "git+https://github.com/dawiegriesel/mcp-dev.git",
        "project-scaffold"
      ]
    }
  }
}

Other MCP clients

Run the server with stdio transport (the default):

project-scaffold

Or without installing:

uvx --from "git+https://github.com/dawiegriesel/mcp-dev.git" project-scaffold

MCP tools

The server exposes five tools via MCP:

list_templates

Returns all available project template combinations and their default stack.

create_project

Scaffolds a complete project. Configuration options:

Parameter

Required

Default

Description

name

yes

—

Project name (lowercase, hyphens allowed)

project_type

yes

—

"work" (Azure) or "personal" (Render.com)

frontend_type

yes

—

"react" (SPA) or "htmx" (server-rendered)

output_dir

yes

—

Parent directory for the project folder

description

no

""

Short project description

db_name

no

derived from name

Database name

include_auth

no

true

Include JWT auth boilerplate

include_alembic

no

true

Include Alembic migrations

api_port

no

8000

Local dev API port

frontend_port

no

3000

Local dev frontend port (React only)

add_component

Adds a component to an existing scaffolded project. Supported components:

  • api_router — FastAPI router with full CRUD endpoints, plus auto-generated model and schema if fields are provided

  • db_model — SQLAlchemy model with matching Pydantic schemas

Components not yet implemented: frontend_page, github_action, docker_service.

validate_project

Checks that a scaffolded project has all expected files. Reports any missing files.

get_project_info

Reads the .scaffold.json metadata from an existing project and returns its configuration.

Example prompts

Once the MCP server is installed, you can use natural language to scaffold and extend projects. Here are some example prompts to get started:

Scaffold a new project

Create a new personal project called "recipe-box" with an HTMX frontend. Put it in ~/projects. Include auth and migrations.

Scaffold a work project named "inventory-tracker" with a React frontend in ~/work. No auth needed but include Alembic.

I want to start a new side project called "budget-app". It should be a simple server-rendered app I can deploy to Render. Create it in ~/dev.

What project templates are available?

Add components to an existing project

Add a "products" API router to my project at ~/projects/recipe-box with fields: title (str), prep_time (int), servings (int), instructions (text).

Add a database model called "category" with fields name (str) and description (text) to ~/projects/recipe-box.

I need a new "orders" endpoint in my inventory-tracker project. It should have fields for customer_name (str), total (float), and fulfilled (bool).

Inspect and validate projects

Check if my project at ~/projects/recipe-box has all the expected files.

What's the configuration for the project in ~/work/inventory-tracker?

Validate the project structure at ~/dev/budget-app and tell me if anything is missing.

Project structure

React projects (multi-repo)

my-app/
├── Makefile
├── README.md
├── .env.example
├── docker-compose.yml
├── docker-compose.override.yml
├── .scaffold.json
├── my-app-api/
│   ├── pyproject.toml
│   ├── Dockerfile
│   ├── app/
│   │   ├── main.py
│   │   ├── config.py
│   │   ├── database.py
│   │   ├── models/
│   │   ├── schemas/
│   │   ├── routers/
│   │   └── auth/          (optional)
│   ├── alembic/            (optional)
│   ├── tests/
│   └── .github/workflows/
└── my-app-frontend/
    ├── package.json
    ├── vite.config.ts
    ├── Dockerfile
    ├── nginx.conf
    ├── src/
    │   ├── App.tsx
    │   ├── api/
    │   ├── pages/
    │   ├── components/
    │   └── styles/
    └── .github/workflows/

HTMX projects (single repo)

my-app/
├── Makefile
├── README.md
├── .env.example
├── docker-compose.yml
├── pyproject.toml
├── Dockerfile
├── .scaffold.json
├── app/
│   ├── main.py
│   ├── config.py
│   ├── database.py
│   ├── models/
│   ├── schemas/
│   ├── routers/
│   ├── templates/
│   │   ├── base.html
│   │   ├── index.html
│   │   └── partials/
│   ├── static/css/
│   └── auth/              (optional)
├── alembic/                (optional)
└── tests/

Development

# Lint
uv run ruff check src/

# Format
uv run ruff format src/

# Test
uv run pytest

Automated documentation updates

This repo includes a docs-pre-commit Claude agent (.claude/agents/docs-pre-commit.md). When working in Claude Code, it triggers automatically before commits to update CHANGELOG.md, README.md, and CLAUDE.md to reflect the changes being committed.

To trigger it manually, tell Claude: "Update the documentation before I commit."

Architecture

src/project_scaffold/
├── __main__.py          # Entry point — runs mcp.run()
├── server.py            # FastMCP server with 5 tool definitions + main()
├── models.py            # Pydantic models (ProjectConfig, ComponentConfig)
├── config.py            # Constants, type maps, template paths
├── generator.py         # Core engine — renders Jinja2 templates to disk
├── renderers/
│   └── api.py           # Component generators (add_router, add_model)
└── templates/           # Jinja2 templates (included in package)
    ├── api/             # FastAPI backend
    ├── frontend/react/  # React + Vite + TypeScript
    ├── frontend/htmx/   # HTMX + Tailwind
    ├── docker/          # Docker Compose
    ├── cicd/azure/      # GitHub Actions for Azure
    ├── cicd/render/     # Render.com blueprint
    ├── alembic/         # Database migrations
    └── common/          # README, Makefile, .env, .gitignore

Available Tools

6 tools
add_componentA

Add a new component to an existing scaffolded project. Supports adding API routers, database models with matching schemas, frontend pages, GitHub Action workflows, or Docker services.

Inputs: config (required). Returns: object

configure_projectA

Returns structured questions for configuring a new project. Call this FIRST when a user wants to create a project. Present the questions to the user, collect their answers, then call create_project with the full configuration.

Inputs: no named parameters. Returns: object

create_projectA

Scaffold a complete new project with API, frontend, database config, Docker setup, and CI/CD pipeline. Returns a summary of created files and next steps.

Set use_current_dir=True to scaffold directly into output_dir instead of creating a new subdirectory.

Inputs: config (required). Returns: object

get_project_infoA

Read an existing scaffolded project and return its configuration, including project type, frontend type, deployment target, and list of components.

Inputs: project_dir (required). Returns: object

list_templatesA

List all available project template combinations and their descriptions.

Inputs: no named parameters. Returns: object

validate_projectA

Validate that a scaffolded project has all expected files and correct structure. Reports missing files and suggests fixes.

Inputs: project_dir (required). Returns: object

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.2/5.0

    Scored across 6 tools

    Disambiguation5/5

    Each tool targets a distinct phase or aspect of project scaffolding: configure_project gathers requirements, list_templates shows options, create_project generates the project, add_component extends it, validate_project checks structure, and get_project_info inspects configuration. There is no overlap or ambiguity between them.

    Naming Consistency5/5

    All tool names follow a consistent verb_noun snake_case pattern: configure_project, list_templates, create_project, add_component, validate_project, get_project_info. This makes the tool set predictable and easy to navigate.

    Tool Count5/5

    With 6 tools, the server is well-scoped for its purpose. Each tool covers a necessary step in the scaffolding workflow without redundancy or bloat.

    Completeness5/5

    The tool set covers the full scaffold lifecycle: template discovery, project configuration, creation, component addition, validation, and inspection. While update/delete operations are missing, they are not expected for a scaffold generator, making the surface complete for its domain.

    Maintenance

    ActivityInactive
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      Not graded
      quality
      D
      maintenance
      A production-ready MCP server scaffold that features built-in authentication, Docker support, and a comprehensive CI/CD release pipeline. It provides a standardized template for deploying servers with multi-transport support and configurable read-only modes.
      MIT
    • A
      license
      A
      quality
      D
      maintenance
      A unified MCP server for scaffolding project structures across multiple frameworks including Spring Boot, React, Vue, Next.js, FastAPI, Django, Flask, Express, and more.
      2
      17 npm
      1
      MIT