Model Context Protocol
An MCP server that serves canonical hadith text —Arabic & English —to AI assistants so they cite from real data instead of quoting from memory.
A narration lookup, a thematic search, a cross-reference check — anything.
It retrieves verified text, narrators, and metadata from a normalized database derived from the hadith-json dataset (sourced from Sunnah.com).
The AI cites exact collection names, hadith numbers, Arabic text, and English translations — backed by the data it fetched, not model memory.
The MCP server exposes five tools. The AI picks the right one based on what you ask.
Returns citation guidance and limitations once per session.
Re-calls get a short repeat unless force_full=True.
Lists every collection: slug, English & Arabic names, and hadith counts.
Fetch by global hadith_id or by collection +
hadith_number. Supports ranges like "1-5"
and optional cross-references.
Semantic (embeddings, default), keyword (SQL substring), or both. Falls back to keyword on rate limits or quota errors.
Algorithmic similarity matches across collections for a given hadith — useful for mutaba'ah-style exploration.
https://hadith-mcp.org
Connect this server to any AI client that supports MCP. One-time setup, then it just works.
Requires a Claude subscription (Pro, Team, or Enterprise).
Go to Settings → Connectors → Customize panel.
Click the + button, then select Add custom connector.
hadithhttps://hadith-mcp.orgOpen the connector and set every tool to Always allow. They're all read-only.
Open a new chat, enable the hadith connector, and ask your question. Look for tool-call indicators in the response.
Requires a ChatGPT subscription. Set up on Web or Desktop once and it syncs to Mobile.
Go to Settings → Apps → Advanced at the bottom.
Turn on Developer mode. A Create app button appears in the top-right.
hadithhttps://hadith-mcp.orgnoneInvoke with @hadith in any chat. Look for Called tool indicators in the response.
Add the remote MCP server in Cursor's settings.
Go to Settings → MCP → Add new MCP server.
streamable-httphttps://hadith-mcp.orgThe hadith tools are now available. Ask about hadith and the agent will call the server automatically.
Any MCP-compatible client can connect. Use the server name hadith, point it at the URL, and keep transport on streamable-http.
hadithhttps://hadith-mcp.orgstreamable-http{
"name": "hadith",
"url": "https://hadith-mcp.org",
"transport": "streamable-http"
}Add to claude_desktop_config.json:
{
"mcpServers": {
"hadith": {
"url": "https://hadith-mcp.org",
"transport": "streamable-http"
}
}
}All text comes from hadith.db, built from the community hadith-json dataset (Sunnah.com). Prefer tool output over model paraphrase.
Always cite with collection slug and id_in_book. Never invent numbering.
Algorithmic matches (embedding similarity + narrator hints) — not a substitute for classical mutaba'ah scholarship.
Labels like muttafaq_alayh or corroborated are
heuristic pipeline tags, not theological rulings.
English translations may be incomplete. Arabic strings are authoritative for the stored row.
Semantic search uses OpenAI embeddings. Keyword search is SQL substring. Neither ranks by scholarly importance.
This tool retrieves hadith text from a database. It is not a scholar, a mufti, or a substitute for qualified Islamic scholarship. AI can still misinterpret context, conflate narrations, or present text without the nuance that trained scholars provide.
Do not blindly trust any AI output on religious matters — whether from this tool or any other. Always verify hadith citations against recognized scholarly sources. Consult qualified scholars for matters of fiqh, aqeedah, and religious practice.
Cross-references and provenance tags in this tool are algorithmic approximations, not the product of classical hadith criticism (ilm al-rijal). They may contain errors. The grading, authenticity, and contextual application of hadith require expertise that no software can replace.