Engineering Documentation MCP Server

Turn any software repository into professional engineering documentation — automatically.

A production-ready Model Context Protocol (MCP) server that analyzes code repositories and publishes comprehensive documentation directly to Google Docs. Works with Claude Code, Cursor, and any MCP-compatible AI assistant.

What It Does

Say this to your AI assistant:

"Document this project."

The server will:

1. Analyze the repository — languages, frameworks, API routes, DB models, env vars, CI/CD, dependencies

2. Generate 13 documentation sections with Mermaid architecture diagrams

3. Publish a formatted Google Doc and return the URL

4. Export to Markdown, HTML, or PDF (optional)

Related MCP server: Documentation MCP Server

Features

Quick Start

1. Clone & Install

git clone https://github.com/your-org/engineering-docs-mcp.git
cd engineering-docs-mcp

# Using uv (recommended)
pip install uv
uv sync

# Or using pip
python -m venv .venv
source .venv/bin/activate  # Windows: .venv\Scripts\activate
pip install -e ".[dev]"

2. Set Up Google OAuth

Note (2026 UI): Google renamed "OAuth consent screen" to Google Auth Platform and reorganized it into tabs: Branding, Audience, Data Access, and Clients. The steps below reflect the current UI.

Step 1 — Create a Google Cloud Project

1. Go to Google Cloud Console

2. Click the project dropdown (top-left) → New Project

3. Give it a name (e.g. Engineering Docs MCP) → Create

Step 2 — Enable the Required APIs

1. Go to APIs & Services → Library

2. Search "Google Docs API" → click it → Enable

3. Search "Google Drive API" → click it → Enable

Step 3 — Configure Google Auth Platform

1. Go to APIs & Services → Google Auth Platform

2. If you see a "Get Started" button, click it — this launches the setup wizard:

   Wizard — App Information

   Field	Value
   App name	Engineering Docs MCP
   User support email	Your Gmail address

   Click Next

   Wizard — Audience

   • Select External (works with any Gmail; choose Internal only for Google Workspace orgs)

   • Click Next

   ⚠️ You cannot change External → Internal later without creating a new project.

   Wizard — Contact Information

   • Enter your email for Google notifications

   • Click Next → Create

Step 4 — Add Scopes (Data Access Tab)

1. Click the Data Access tab

2. Click Add or Remove Scopes

3. In the filter/search box:

   • Search Google Docs → check /auth/documents

   • Search Google Drive → check /auth/drive

4. Click Update → Save

Step 5 — Add Test Users (Audience Tab)

1. Click the Audience tab

2. Scroll down to Test users → click + Add Users

3. Enter your Gmail address → Add → Save

While in Testing mode only listed emails can authenticate. You never need to publish the app for personal/dev use.

Step 6 — Create OAuth Credentials (Clients Tab)

1. Click the Clients tab

2. Click + Create Client

3. Application type: Desktop app

4. Name: Engineering Docs MCP → Create

5. In the dialog that appears, click Download JSON

6. Rename the downloaded file to credentials.json

7. Place it in the project root:

documentation_mcp/
├── credentials.json   ← here
├── server.py
└── ...

Quick Reference — Old vs New UI

3. Configure Environment

cp .env.example .env
# Edit .env if needed — defaults work for most setups

4. Run the Server

python server.py

On first run, a browser window will open for Google OAuth authentication. After authorizing, a token.json is saved for subsequent runs.

Connecting to Claude Code

Add the server to your Claude Code MCP configuration:

# Edit ~/.claude/claude_desktop_config.json (Mac/Linux)
# or %APPDATA%\Claude\claude_desktop_config.json (Windows)

{
  "mcpServers": {
    "engineering-docs-mcp": {
      "command": "python",
      "args": ["/absolute/path/to/engineering-docs-mcp/server.py"],
      "env": {
        "GOOGLE_CREDENTIALS_FILE": "/absolute/path/to/credentials.json"
      }
    }
  }
}

Or using uv:

{
  "mcpServers": {
    "engineering-docs-mcp": {
      "command": "uv",
      "args": [
        "--directory", "/absolute/path/to/engineering-docs-mcp",
        "run", "python", "server.py"
      ]
    }
  }
}

Restart Claude Code to load the server.

Connecting to Cursor

Add to Cursor's MCP settings (~/.cursor/mcp.json):

{
  "mcpServers": {
    "engineering-docs-mcp": {
      "command": "python",
      "args": ["/absolute/path/to/engineering-docs-mcp/server.py"]
    }
  }
}

Example Prompts

Once connected, use natural language:

Document this project.

Analyze /path/to/my-api and create a Google Doc. Share it with team@example.com.

Generate API documentation for the repository at /home/user/projects/my-service.

Export the documentation for /path/to/repo as Markdown.

Show me the database schema for /path/to/project.

Generate a README for /path/to/my-project.

Update Google Doc abc123 with the latest docs for /path/to/repo.

Available MCP Tools

Generated Documentation Sections

Every generated document includes:

1. Executive Summary — What the project does in plain language

2. Project Overview — Purpose, features, stats, contributors

3. Technology Stack — Languages, frameworks, databases, infrastructure

4. Folder Structure — Annotated directory tree

5. Architecture — Patterns, request flow, Mermaid flowchart

6. API Documentation — All endpoints with methods, auth, handlers

7. Database Documentation — Models, fields, relationships, ER diagram

8. Environment Variables — All env vars with descriptions and examples

9. Installation Guide — Step-by-step setup instructions

10. Deployment — Docker, Kubernetes, CI/CD instructions

11. Dependencies — Production and dev dependencies with purposes

12. Security Notes — Auth, secrets, recommendations

13. Future Improvements — Suggested enhancements based on analysis

Docker

# Copy environment file
cp .env.example .env

# Build and run
docker compose up -d

# View logs
docker compose logs -f mcp-server

The Docker setup mounts:

• ./credentials.json — Google OAuth credentials (read-only)

• ./token.json — OAuth token (persisted across restarts)

• ./output/ — Generated export files

• ${REPOS_ROOT:-/repos} — Host directory for repository access

Development

# Install dev dependencies
uv sync

# Run tests
pytest

# Run with coverage
pytest --cov=. --cov-report=html

# Lint
ruff check .
ruff format .

# Type check
mypy .

Project Structure

engineering-docs-mcp/
│
├── server.py                 # FastMCP server with 12 tools
│
├── analyzers/                # Repository analysis engines
│   ├── orchestrator.py       # Runs all analyzers in sequence
│   ├── structure.py          # File/folder structure + language detection
│   ├── framework.py          # Framework, CI/CD, auth, deployment detection
│   ├── api.py                # API endpoint extraction
│   ├── database.py           # ORM model extraction
│   ├── environment.py        # Environment variable extraction
│   ├── dependencies.py       # Dependency manifest parsing
│   └── git_info.py           # Git metadata
│
├── generators/               # Documentation section generators
│   ├── orchestrator.py       # Assembles complete documentation
│   ├── overview.py           # Executive summary + project overview
│   ├── tech_stack.py         # Technology stack + architecture + Mermaid diagrams
│   ├── structure.py          # Folder structure section
│   ├── api_docs.py           # API documentation tables
│   ├── database_docs.py      # Database documentation tables
│   ├── env_docs.py           # Environment variables table
│   ├── installation.py       # Installation guide + deployment
│   ├── security.py           # Security notes + future improvements
│   └── dependencies_docs.py  # Dependencies tables
│
├── google_docs/              # Google Docs integration
│   ├── auth.py               # OAuth 2.0 authentication
│   ├── client.py             # Create, update, share documents
│   └── formatter.py          # batchUpdate request builder
│
├── services/
│   ├── documentation.py      # High-level pipeline orchestrator
│   └── export.py             # Markdown / HTML / PDF export
│
├── models/
│   ├── analysis.py           # Pydantic models for repository analysis
│   └── documentation.py      # Pydantic models for documentation
│
├── config/
│   └── settings.py           # Pydantic Settings configuration
│
├── tests/                    # pytest test suite
│
├── Dockerfile
├── docker-compose.yml
├── pyproject.toml
└── .env.example

Supported Frameworks

Environment Variables

Extending the Server

Adding a New Framework Detector

Edit analyzers/framework.py and add entries to FRAMEWORK_SIGNATURES:

Framework.MY_FRAMEWORK: [
    ("*.py", "from my_framework import"),
    ("requirements*.txt", "my-framework"),
],

Add the framework to the Framework enum in models/analysis.py.

Adding a New Documentation Section

1. Create a generator in generators/my_section.py extending BaseGenerator

2. Register it in generators/orchestrator.py

3. Add the field to ProjectDocumentation in models/documentation.py

Troubleshooting

FileNotFoundError: credentials.json Download OAuth credentials from the Clients tab in APIs & Services → Google Auth Platform and place credentials.json in the project root.

Can't find "OAuth consent screen" in the sidebar Google renamed it to Google Auth Platform. Navigate to APIs & Services → Google Auth Platform instead.

access_blocked error during OAuth login` Your email isn't in the Test Users list. Go to Google Auth Platform → Audience tab → Test users and add your Gmail address.

Token refresh failed Delete token.json and restart the server. A new browser authentication window will open.

Scopes not showing in Data Access tab The APIs must be enabled first. Go to APIs & Services → Library, enable Google Docs API and Google Drive API, then return to the Data Access tab.

PDF export fails WeasyPrint requires system libraries. On Ubuntu/Debian:

apt-get install libpango-1.0-0 libharfbuzz0b libpangoft2-1.0-0

No API endpoints detected Ensure the repository uses supported frameworks. Currently supports FastAPI, Flask, Django, Express, and Next.js route handlers.

License

MIT — see LICENSE for details.