Seerah Knowledge Base 📖

Ask any question about the life of the Prophet Muhammad ﷺ and get answers grounded in 4 scholarly books — fully offline, no API key, free.

A searchable knowledge base of the Seerah (Prophetic biography) that plugs directly into Claude Code or Claude Desktop as an MCP server, or runs standalone from the command line. Semantic search runs locally using the all-MiniLM-L6-v2 embedding model — your questions never leave your machine and there is nothing to pay for.

The books

1,103 embedded passages across four widely-respected works:

Every search result includes a Source: line naming the book it came from, so answers are always attributable.

Quick start

Requires Python 3.10+.

git clone https://github.com/Fezicles/seerah-knowledge-base.git
cd seerah-knowledge-base

# Mac / Linux
bash setup.sh

# Windows: double-click setup.bat

Setup creates a virtual environment, installs dependencies, and writes a .mcp.json so Claude Code picks up the search_seerah tool automatically. Then just run:

claude

…from this folder and ask anything:

"What happened at the Treaty of Hudaybiyyah, and why did some companions find it difficult?"

"Tell me about Khadijah (RA) and her role in the early revelation."

"Compare how the books describe the Battle of Uhud."

Claude searches the knowledge base, reads the passages, and synthesises an answer with book citations.

Use it without Claude (standalone CLI)

.venv/bin/python search.py "Who was Salman al-Farisi?" --results 6

Prints the most relevant passages with relevance scores and sources.

Use it with Claude Desktop

Add to claude_desktop_config.json (path shown for macOS: ~/Library/Application Support/Claude/):

{
  "mcpServers": {
    "seerah-kb": {
      "command": "/path/to/seerah-knowledge-base/.venv/bin/python",
      "args": ["/path/to/seerah-knowledge-base/mcp_server.py"]
    }
  }
}

How it works

• The four books were chunked into 1,103 passages and embedded with all-MiniLM-L6-v2 (a small, fast sentence-transformer). The pre-built index ships in this repo (lightrag_store/vdb_chunks.json, ~9 MB) — you never need to rebuild it.

• At query time, your question is embedded locally and matched against the index by cosine similarity. Top passages are returned as plain text for the model (or you) to read.

• No API key, no cloud calls. The only download is the embedding model itself (~90 MB from Hugging Face, once, on first search — cached after that). After that it works fully offline.

Repository layout

mcp_server.py       MCP server exposing the search_seerah tool
search.py           Standalone CLI search
setup.sh / .bat     One-time setup (venv + deps + MCP config)
configure_mcp.py    Writes .mcp.json with this machine's paths
lightrag_store/     Pre-built embedding index (1,103 chunks) + full source
                    text (kv_store_full_docs.json — useful if you want to
                    re-chunk or rebuild with a different embedding model)
CLAUDE.md           Instructions telling Claude Code how to use the KB

Credits

Built by @hamismahmood9.

Content & license

The code in this repository is MIT-licensed (see LICENSE).

The indexed passages are excerpts from published books whose copyrights belong to their respective authors and publishers. They are included here solely to enable personal study and educational research of the Seerah. If you benefit from this tool, please support the authors — buy the books. If you are a rights holder and would like content removed, open an issue and it will be handled promptly.

May it be a means of benefit. Corrections and contributions welcome.