古兰经参考MCP服务器

Quran Tafseer MCP Server

v1.0.0

A neutral Quran reference MCP server written in Free Pascal. Provides side-by-side comparison of Arabic base text and English translations over the MCP stdio protocol. The server provides materials only — no interpretation or commentary.

Features

• 6 bundled public-domain translations (Palmer, Rodwell, Sale, Yusuf Ali, Pickthall, Shakir)
• Arabic base text (Uthmani script, Hafs reading) bundled and installed automatically
• 50+ downloadable translations from quran-api catalog
• Full-text search across all installed translations (SQLite FTS5)
• Reference resolution — accepts 2:255, Q 2:255, Al-Baqarah 255
• Two output modes — structured JSON or terminal-friendly preformatted text
• Zero network calls at runtime — all data is local; downloads only during setup

MCP Tools

All tools accept format: "terminal" for preformatted text output (default: "structured" JSON).

Installation

Pre-built binaries for Windows, Linux, and macOS are available on the Releases page.

Building from source

This project is written in Free Pascal and uses the Lazarus build tool lazbuild.

Windows

Install Free Pascal and Lazarus using the combined installer from SourceForge. Choose the latest Lazarus release (3.6+), which bundles Free Pascal 3.2.2. After install, ensure lazbuild is on your PATH (typically C:\Lazarus\lazbuild.exe).

lazbuild quran-tafseer-mcp.lpi

Linux (Debian/Ubuntu)

sudo apt-get update
sudo apt-get install -y fpc lazarus libsqlite3-0
lazbuild quran-tafseer-mcp.lpi

macOS

brew install fpc lazarus sqlite
lazbuild quran-tafseer-mcp.lpi

Runtime dependency

sqlite3 — required for full-text search. On Windows, place sqlite3.dll alongside the executable (download from sqlite.org). On Linux/macOS it's typically already installed; if not, use apt install libsqlite3-0 or brew install sqlite.

Quick Start

# First-run setup: install bundled translations only (no network)
quran-tafseer-mcp init --bundled-only

# First-run setup: install bundled + download all catalog translations
quran-tafseer-mcp init --all

# Run as MCP server
quran-tafseer-mcp mcp

All commands use a platform-default data root when --data is omitted:

• Windows: %LOCALAPPDATA%\quran-tafseer-mcp
• Linux/macOS: $XDG_DATA_HOME/quran-tafseer-mcp (or ~/.local/share/quran-tafseer-mcp)

Override with --data <path> if you want a custom location.

Client Setup

Since --data is optional (the server uses a platform-default data root), client configuration is simple — no need to hardcode a data path.

Claude Code

Option 1: CLI (recommended)

claude mcp add quran-tafseer-mcp -- /path/to/quran-tafseer-mcp mcp

Option 2: Settings file (project .mcp.json)

{
  "mcpServers": {
    "quran-tafseer-mcp": {
      "command": "/path/to/quran-tafseer-mcp",
      "args": ["mcp"]
    }
  }
}

Windows example:

{
  "mcpServers": {
    "quran-tafseer-mcp": {
      "command": "C:/path/to/quran-tafseer-mcp.exe",
      "args": ["mcp"]
    }
  }
}

To use a custom data root, add "--data", "/path/to/data" to the args array.

Codex

codex mcp add quran-tafseer-mcp -- /path/to/quran-tafseer-mcp mcp

Or in ~/.codex/config.toml:

[mcp_servers.quran-tafseer-mcp]
command = "/path/to/quran-tafseer-mcp"
args = ["mcp"]

First run via MCP

On first MCP connection with an empty data root, the server automatically installs the 7 bundled corpora (6 English translations + Arabic base text). To download additional translations, ask your AI assistant to call quran.setup.

CLI Commands

All commands use the platform-default data root when --data is omitted.

# MCP server
quran-tafseer-mcp mcp [--data <path>] [--log-level error|warn|info|debug]

# First-run setup
quran-tafseer-mcp init [--data <path>]                 # interactive setup
quran-tafseer-mcp init [--data <path>] --bundled-only  # bundled corpora only (no network)
quran-tafseer-mcp init [--data <path>] --all           # bundled + download all catalog translations

# Corpus management
quran-tafseer-mcp corpus list [--data <path>]
quran-tafseer-mcp corpus validate [--data <path>] [--id <id>]
quran-tafseer-mcp corpus add [--data <path>] --id <id> --file <path> --format <fmt> --title "..."

# Search index management
quran-tafseer-mcp index build [--data <path>] [--id <id> | --all]

# Other
quran-tafseer-mcp --version
quran-tafseer-mcp --help

Output Format

All tools accept a format parameter:

• structured (default) — JSON objects with full data and citations
• terminal — preformatted text blocks optimized for terminal display. Arabic text is placed on its own lines to avoid BiDi rendering issues.

Example terminal output for quran.get_ayah:

-- Q 1:1 --

[Arabic · ar.uthmani]
بِسْمِ اللَّهِ الرَّحْمَٰنِ الرَّحِيمِ

[en.yusufali.1934]
In the name of Allah, Most Gracious, Most Merciful.

[en.pickthall.1930]
In the name of Allah, the Beneficent, the Merciful.

Data Layout

quran-tafseer-mcp-data/
  corpora/quran/
    ar.uthmani/original.tsv + manifest.json
    en.palmer.1880/original.tsv + manifest.json
    ...
  indexes/quran/
    ar.uthmani.sqlite
    en.palmer.1880.sqlite
    ...

Supported Corpus Formats

Credits

• Quran API
• Tanzil
• Quranic Universal Library
• QuranEnc.com
• Nurul Hanisah Kamarul Zaman

License

Code: GPL-3.0

Bundled corpora: Public domain (Wikisource) and Project Gutenberg license. See individual corpus directories for details.

Users are responsible for ensuring they have the right to use any additional translations they import.