ataturk-mcp 0.1.0

An MCP server that exposes the corpus of Atatürk's speeches, statements, telegrams and proclamations (1906-1938) for researchers worldwide. Created by bugraayan.com.

Atatürk MCP

Created by Buğra Ayan — bugraayan.com

Tüm dünyadan Atatürk üzerine araştırma yapanlar için açık kaynak bir köprü.

A Model Context Protocol server that exposes the complete corpus of Mustafa Kemal Atatürk's speeches, statements, telegrams and proclamations (1906-1938) to any MCP-aware AI client (Claude Desktop, Cursor, Cline, Windsurf, …).

Built for researchers, historians, journalists and students anywhere in the world who want to ask LLMs questions like:

• "What did Atatürk say about women's rights in 1923?"
• "Find quotes about education and modernisation."
• "Show me his opening address to the Grand National Assembly on 1 March 1922."
• "Compare the 1927 Nutuk's treatment of the War of Independence with his 1933 10th Year Speech."

Corpus

Source	Coverage	Format	Speeches
ATAM — Atatürk Araştırma Merkezi "Söylev ve Demeçleri" Cilt I-III (2024 edition)	1906-1938, the definitive corpus	PDF	366
Vikikaynak	Individual speeches, telegrams, all TBMM opening addresses (1920-1938)	HTML	45
Internet Archive — Nutuk (English)	The 1927 Nutuk in English	PDF / DjVu	20 chapters
Total	1906-1938	SQLite + FTS5	411 speeches + Nutuk

Atatürk passed away in 1938; his words are in the public domain. Editorial annotations from the ATAM edition are not redistributed — only the speech bodies themselves, with a source_ref pointing back to the corresponding ATAM volume and page number for academic citation.

Quick start

1. Install

From PyPI (recommended for end users):

pip install ataturk-mcp                # runtime only
pip install "ataturk-mcp[etl]"         # also includes the ETL scripts

From source (for hacking on the ETL):

git clone https://github.com/bugraayan/ataturk-mcp.git
cd ataturk-mcp
python3 -m venv .venv && source .venv/bin/activate
pip install -e ".[etl]"

2. Build the database (one-off, ~1 minute)

python scripts/fetch_atam.py            # ~10 MB of PDFs from atam.gov.tr
python scripts/fetch_wikisource.py      # MediaWiki API, ~45 pages
python scripts/fetch_nutuk_en.py        # ~2 MB DjVu text from Internet Archive
python scripts/build_db.py              # produces data/speeches.db (~11 MB)

Or, if you only want the core corpus:

python scripts/fetch_atam.py
python scripts/build_db.py --skip-wikisource --skip-nutuk

A pre-built speeches.db is also published on the GitHub Releases page so end users do not need to run the ETL themselves.

3. Run the MCP server

ataturk-mcp                  # stdio transport (default, used by all MCP clients)
# or
fastmcp run src/ataturk_mcp/server.py:mcp
# or
python -m ataturk_mcp.server

Connecting to MCP clients

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or the equivalent on your platform and add:

{
  "mcpServers": {
    "ataturk": {
      "command": "/absolute/path/to/.venv/bin/ataturk-mcp"
    }
  }
}

If your database lives somewhere other than the repo, set the path explicitly:

{
  "mcpServers": {
    "ataturk": {
      "command": "/absolute/path/to/.venv/bin/ataturk-mcp",
      "env": {
        "ATATURK_MCP_DB": "/path/to/speeches.db"
      }
    }
  }
}

Restart Claude Desktop; the hammer icon now exposes the Atatürk tools.

Cursor

Edit ~/.cursor/mcp.json (or the project-level .cursor/mcp.json):

{
  "mcpServers": {
    "ataturk": {
      "command": "/absolute/path/to/.venv/bin/ataturk-mcp"
    }
  }
}

Cline / Continue / any MCP host

Use the same command line with whichever JSON the host expects. The server speaks stdio by default and follows the MCP 2024-11-05 spec.

Tools exposed

Tool	Purpose
search_speeches(query, lang, year_from, year_to, kind, limit)	Full-text BM25 search over the entire corpus, with snippets. Diacritic-insensitive; supports FTS5 operators (AND, OR, NEAR, "phrases", prefix*).
get_speech(speech_id, lang)	Return one speech in full, in Turkish or English.
list_speeches(year, year_from, year_to, kind, source, limit, offset)	Browse the corpus chronologically.
random_speech(lang, kind)	Pick a random speech (useful for daily-quote agents).
list_topics() / speeches_by_topic(topic_id, limit)	Topical browsing (when ATAM Konular İndeksi is loaded).
nutuk_search(query, lang, limit)	Search within Nutuk (1927).
nutuk_chapter(chapter, lang)	Return one Nutuk chapter (1-20).
cite(speech_id)	Generate APA / MLA / Chicago citations.
corpus_stats()	Summary statistics about the corpus.

Resources

URI	Description
ataturk://speech/{speech_id}	Plain-text rendering of a single speech with header.
ataturk://nutuk/{chapter}/{lang}	One Nutuk chapter.
ataturk://corpus/stats	Statistics as JSON.

Prompts

Prompt	Use
analyze_speech(speech_id)	Scholarly analysis template (context, rhetoric, themes, citation).
find_quote(theme, n_quotes)	Theme-based quote hunter across the corpus.

Development

pip install -e ".[etl,dev]"
pytest -q
ruff check .

The test suite uses an in-memory seeded SQLite DB and FastMCP's in-process client, so it runs in under a second and does not require the production DB to be built.

Architecture

ATAM PDFs ─┐
Vikikaynak ─┼─► ETL scripts ─► SQLite (FTS5) ─► FastMCP stdio server ─► AI clients
Nutuk EN ──┘                  speeches.db

• ETL (scripts/) is fully decoupled from the server (src/ataturk_mcp/).
• The server opens the DB read-only and is safe to run in parallel from multiple clients.
• Turkish search quality: FTS5 with unicode61 remove_diacritics 2 plus application-level I/İ normalisation in db.normalise_query.

Publishing to PyPI

The project is wired to publish from GitHub Actions (see .github/workflows/release.yml) when you push a tag of the form vX.Y.Z. For manual publishing:

pip install build twine
python -m build                          # produces dist/*.whl and dist/*.tar.gz
twine check dist/*
twine upload dist/*                      # uploads to https://pypi.org/project/ataturk-mcp/
# or for a dry run on TestPyPI:
twine upload --repository testpypi dist/*

Credits & author

• Author: Buğra Ayan — bugraayan.com
• Email: [e-mail]
• Atatürk Araştırma Merkezi (ATAM) — for digitising and editing the Söylev ve Demeçleri corpus, the gold-standard source for this project.
• Vikikaynak / Türkçe Wikisource contributors — for transcribing individual addresses and the TBMM opening speeches.
• Internet Archive — for hosting the public-domain English Nutuk.

If you build research or journalism on top of this server, please cite both Atatürk's words (via the in-tool cite command) and the original sources (ATAM volume + page, or the Wikisource URL).

License

MIT for the code (© Buğra Ayan / bugraayan.com). The speech texts themselves are in the public domain (Atatürk died in 1938). ATAM, TBMM and Wikisource are credited as the digital sources used to build the corpus; please respect each source's terms when redistributing.