Tafsir Center for Quranic Studies has launched Tafsir MCP, a scholarly Model Context Protocol server for the Holy Quran in the age of artificial intelligence. Built on Anthropic's MCP protocol, it gives any compatible AI assistant — Claude Code, Claude Desktop, VS Code, and others — certified scholarly access to Quranic text, classical commentaries, and linguistic analysis, while eliminating the hallucinations that language models produce when handling Quranic content.
All data is reviewed and certified by the Center's researchers. The database is opened read-only with strict surah/ayah boundary validation via Pydantic v2, preventing any text generation outside verified sources.
Scholarly Content
- 6,236 ayahs in Uthmani script.
- 8 tafsir sources: Jami' al-Bayan by al-Tabari (d. 310H), Tafsir Ibn Kathir (d. 774H), Ma'alim al-Tanzil by al-Baghawi (d. 510H), Taysir al-Karim al-Rahman by al-Sa'di (d. 1376H), al-Tafsir al-Muyassar, and al-Mukhtasar — the latter available in Arabic, English, and Bengali.
- 77,432 words with full linguistic analysis: i'rab, sarf, root, and rasm.
- 1,891 indexed roots searchable with full occurrence statistics.
- Asbab al-nuzool with full isnad for 201 ayahs.
- Qira'at variants per ayah and word.
- Detailed statistics per surah and mushaf page: word count, character count, longest word.
Available Tools (13)
- Ayah: fetch ayah text with optional tajweed/i'rab, retrieve tafsir from up to 8 sources in one call, and get asbab al-nuzool with full isnad.
- Surah: full surah info (goals, virtues, revelation context) and detailed statistics.
- Word: full analysis (i'rab, sarf, meaning, root, qira'at), all occurrences of a root across the Quran, and root statistics.
- Search: fast FTS5 diacritic-free search in Quran text, and LIKE search within any tafsir body.
- General: Quran-wide statistics and page-level fawaed.
Installation
Install for Claude Code with a single command:
claude mcp add tafsir --scope user -- uvx tafsir-mcpOr install directly via pip/uvx:
pip install tafsir-mcpOn first run, the database (~214 MB) is downloaded automatically from Hugging Face and cached at ~/.cache/tafsir-mcp/. The server then works fully offline. A ready-to-use remote endpoint is also available at mcp.tafsir.net/mcp.
License: code is open-source under MIT; Quranic data under CC BY 4.0 with required attribution to Tafsir Center for Quranic Studies.
Full project guide · Setup walkthrough (YouTube) · Source on GitHub · Data on Hugging Face · Contact: [email protected]
