bible-mcp

An MCP server for Christian scholarship and research. Non-commercial, aiming to become a public resource. See corpus-survey.md for the full source/license landscape and ROADMAP.md for direction.

What's in it (v0.12)

All in one SQLite file (db/bible.db, ~370MB) with FTS5 full-text search. Prose works are addressed as WORK.chapter.paragraph. See DESIGN.md for the architecture rationale.

Tools

• get_passage(reference, version="BSB") — Bible text for a reference, e.g. John 3:16, John 3:16-18, Genesis 1. Versions: BSB, WEB (Apocrypha requires version="WEB").
• search(query, version="BSB", book="", limit=20) — Full-text search, stemmed and ranked (BM25). Supports quoted phrases and AND/OR/NOT, e.g. faith AND works NOT law. version also takes any prose work id. Note: stemming conflates related surface forms (e.g. "desert" also matches "deserted") — quote exact phrases or add AND-terms to disambiguate.
• semantic_search(query, top_k=12, kind="", hybrid=True) — Meaning-based search across the whole corpus (scripture + prose), hybrid-fused with keyword search (RRF) by default. Finds passages on a theme even with no shared words, e.g. "divine self-emptying". kind optional: verse | window | paragraph.
• find_similar(reference, top_k=10) — Nearest passages by embedding to a given verse or prose paragraph, across scripture, Apocrypha, and the classics. Powers parallel-finding across corpus layers.
• word_study(query, language="", limit=15) — Original-language word study by Strong's number (zero-padding optional), lemma (pointed or unpointed), Hebrew/Aramaic transliteration (e.g. miqweh, macrons optional), or English gloss. Returns occurrence counts, gloss range, book distribution, a full scholarly lexicon entry when available (BDB/Strong's for Hebrew, Abbott-Smith for Greek — richer than the gloss, truncated to ~700 chars), and sample verses. Lettered homograph variants (e.g. H4723 vs H4723a — same written form, different word) are surfaced together. language optional: grc | hbo | arc.
• get_interlinear(reference) — Word-by-word original language for a verse or short range: surface form, lemma, Strong's, gloss, morphology (Hebrew/Aramaic OT, Greek NT). Versemap-corrected across all books, not just Psalms. For OT verses also shows Septuagint Greek surface text where available (no lemma/Strong's/morphology in that source).
• get_cross_references(reference, limit=20) — Cross-references for a verse (OpenBible.info, ranked by community votes), with the target text included.
• get_citations(reference, limit=20) — Where a verse is cited by name in the patristic corpus, extracted from the translators' own footnotes (tier 1). Complements get_cross_references (scripture→scripture); this is patristic text→scripture. Coverage is sparse by design — an empty result doesn't mean uncited; full-text search within the patristic works is the thorough probe.
• get_entity(name, entity_type="") — Look up a biblical person, place, event, or people group by name (Theographic knowledge graph); returns details and where they appear. entity_type optional: person | place | event | people_group.
• entities_in_passage(reference) — People, places, and events linked to a verse or chapter, e.g. Genesis 14.
• read_work(work, chapter=1, start=1, end=5) — Read a prose work by paragraph range (CONFESSIONS, IMITATION, PILGRIM, PRESENCE, JULIAN, ORTHODOXY, 1CLEMENT, BARNABAS, and others — see corpus_info() for the full list). Use search(version=<WORK>) to find passages first.
• compare_versions(reference) — A verse or short range in BSB and WEB, side by side.
• corpus_info() — What's in the corpus: documents, layers, licenses, and counts.

Prompts

The server also exposes the synthesis pipeline as MCP prompts, generated from .claude/skills/ (the source of truth): corpus_survey(theme) — Layer 0 raw reading discipline; corpus_composer(theme) — Layer 2 research-brief composition. Any MCP client that installs bible-mcp gets the discipline bundled with the data.

Setup

pip install -r requirements.txt   # core: just `mcp`
pip install fastembed numpy       # optional semantic tier (semantic_search, find_similar)
python3 scripts/build_db.py       # rebuild db from data/sources (optional; db ships built)
python3 scripts/embed.py          # regenerate embeddings (optional; ships built; resumable)

The first semantic query downloads the embedding model (~65MB, one time). Without fastembed/numpy installed, all non-semantic tools work normally. See DESIGN.md for architecture decisions (model choice, chunking, hybrid retrieval).

Claude Desktop / Cowork — add to MCP config:

{
  "mcpServers": {
    "bible-mcp": {
      "command": "python3",
      "args": ["/path/to/bible-mcp/server.py"]
    }
  }
}

Layout

data/sources/   raw acquired sources (immutable; re-derive, never edit)
data/versemap.tsv        derived MT/NA <-> English verse alignments
data/additions-*.json    prepared incremental ingestions (applied via scripts/ingest_additions.py)
scripts/        schema.sql, lib_refs.py, build_db.py, build_versemap.py, align_splits.py, ingest_additions.py
db/bible.db     the built corpus (SQLite + FTS5)
outputs/        synthesis-pipeline artifacts (Layer 0 surveys, Layer 2 briefs, reports)
.claude/skills/ corpus-survey + corpus-composer (pipeline skills; source for MCP prompts)
server.py       MCP server (FastMCP, read-only on the db)

Public edition

This is a non-commercial public resource. Licensing is layered — sources keep their own licenses (all PD / CC BY / CC BY-SA); the code is PolyForm Noncommercial 1.0.0; the compilation and derived data (versemap, citation graph, embeddings, outputs layer) are CC BY-NC 4.0. Full details: LICENSE.md, attributions in NOTICE.md.