Docva — Canva Education Document Generator

Headless MCP server that generates teacher verification documents — employment letters, teacher ID cards, teaching licenses, payslips, and more — across 13 countries.

Portable, self-contained, and installable anywhere.

Table of Contents

• Sample outputs

• Features

• Supported countries

• Requirements

• Installation

• Usage — MCP server

• Legacy GUI

• Project structure

• Adding a new country

• License

Related MCP server: office-worker-mcp

Sample outputs

Documents are rendered as high-resolution PNGs. Examples generated by this tool:

Features

• Headless MCP server — document generation exposed as agent-callable tools over stdio.

• 13 countries, each with its own document types and local conventions.

• Real school databases with street addresses, districts, and contact info.

• Consistent profile photos per person — hash-based selection from a bundled, gender-aware photo pool.

• Cross-platform fonts — DejaVu Sans bundled; no system-font dependency.

• Packaged & portable — ships as a self-contained wheel (code + photos + fonts) installable with one command.

Supported countries

Requirements

• Python 3.10+

• Dependencies (installed automatically): Pillow, mcp

Installation

From the built wheel

pip install dist/docva_doc_generator-0.1.0-py3-none-any.whl

From source (editable)

pip install -e .

Via uv

uvx --from . docva-mcp

Usage — MCP server

The server speaks MCP over stdio — the transport used by most agent runtimes (Hermes, Claude Desktop, and any MCP client). Connect it, discover the tools, then call them.

Step 1 — Install & verify

# from the built wheel
pip install dist/docva_doc_generator-0.1.0-py3-none-any.whl

# or editable from source
pip install -e .

Verify the install and that bundled assets resolve:

python -c "from countries.utils import load_font, get_profile_photo; \
print(load_font(30).getname()); print(get_profile_photo((280,340), person_id='x', gender='Male') is not None)"
# ('DejaVu Sans', 'Book')   <-- bundled font, not system
# True                      <-- bundled photo found

Step 2 — Run the server

# After install:
docva-mcp

# Or from source:
python mcp_server.py

It blocks and waits for MCP requests over stdin/stdout — don't run it as a foreground terminal app expecting prompts.

Step 3 — Register in your agent runtime

Point your MCP client at the docva-mcp command:

{
  "mcpServers": {
    "docva": {
      "command": "docva-mcp",
      "args": []
    }
  }
}

If docva-mcp isn't on your PATH, use the absolute path to your interpreter and module instead:

{
  "mcpServers": {
    "docva": {
      "command": "/path/to/python",
      "args": ["-m", "mcp_server"]
    }
  }
}

Tools

list_countries_tool()

No arguments. Returns one result item per country — { code, name, document_types }. (Because a list return is split into one MCP content item per entry, iterate content to see them all.)

list_schools(country: str)

• country (required) — country code from list_countries_tool (e.g. "us").

• Returns one result item per school — { name, address, town, postcode, state, phone, lea }. Iterate content to see them all.

generate_documents(...)

Returns { country, school, document_types, files, count, output_dir } — files are absolute PNG paths.

Connect from a Python client

Minimal working client (requires pip install mcp):

import asyncio
from mcp import ClientSession, StdioServerParameters
from mcp.client.stdio import stdio_client

async def main():
    params = StdioServerParameters(command="docva-mcp", args=[])
    async with stdio_client(params) as (read, write):
        async with ClientSession(read, write) as session:
            await session.initialize()

            countries = await session.call_tool("list_countries_tool", {})
            # A list return is split into one content item per entry:
            for item in countries.content:
                print(item.text)

            res = await session.call_tool("generate_documents", {
                "country": "us",
                "first_name": "John",
                "last_name": "Smith",
                "school_name": "Valley High",
                "position": "Head of Science Department",
                "date_of_birth": "12/05/1988",
                "gender": "Male",
            })
            print(res.content[0].text)

asyncio.run(main())

Typical agent workflow

1. Call list_countries_tool to see what's available.

2. Call list_schools("us") to pick a real school.

3. Call generate_documents(...) with the chosen country, school, and person details.

4. Read the returned PNG paths and use the files.

Generated PNGs are written to output/ (or the output_dir you pass).

Legacy GUI

A tkinter (CustomTkinter) GUI is still available for manual use. The core generation logic is shared.

python main_gui.py        # on Windows, use run.bat (sets TCL_LIBRARY)

The MCP server is the primary, headless interface. The GUI is optional and not required for the skill.

Project structure

docva/
├── countries/            # Document generation core (package)
│   ├── base.py           # CountryGenerator ABC (contract)
│   ├── utils.py          # Fonts, profile photos, shared helpers
│   ├── foto/             # Bundled profile photos (package data)
│   ├── fonts/            # Bundled DejaVu fonts (package data)
│   └── <country>/        # One package per country
├── mcp_server.py         # MCP server exposing tools
├── main_gui.py           # Legacy tkinter GUI
├── docs/examples/        # Sample rendered documents
├── pyproject.toml        # Packaging, deps, entry point
├── output/               # Generated documents (git-ignored)
└── run.bat               # Windows GUI launcher

Adding a new country

1. Create countries/<code>/__init__.py with a class inheriting countries.base.CountryGenerator.

2. Implement the abstract methods: get_country_name, get_country_code, get_schools_data, get_first_names, get_last_names, get_positions, get_document_types, generate_document.

3. Register it in countries/__init__.py via register_country("<code>", <Name>Generator).

4. Optionally add a display label in main_gui.py (get_country_list / on_country_change).

The new country is automatically picked up by the MCP list_countries_tool and list_schools.

License

MIT