Orthotomeo

What orthotomeo MCP server does

The orthotomeo MCP server exposes a read-only scripture-study engine through the Model Context Protocol. It works with a derived SQLite database built from supplied corpus files covering the Greek New Testament, Hebrew Old Testament, Septuagint, and several public-domain English translations. The corpus remains the source of truth; the database is an index that can be rebuilt.

Its purpose is evidence retrieval rather than interpretation. Returned words, verses, lemmas, and citations come from source rows through deterministic queries. An LLM can reason about those results, but the engine does not summarize, paraphrase, select supposedly relevant material, or settle a word’s meaning using a lexicon shortcut.

How it works

The engine keeps each text tradition’s own versification and canon. When editions differ, the difference is surfaced as a caveat rather than being silently forced into a shared numbering scheme. This matters for work involving the Septuagint, deuterocanonical books, or passages whose verse boundaries vary between traditions.

Concordance searches use a complete-or-fail model: a successful query returns every matching occurrence, while an error indicates that the result could not be completed. The design avoids silently returning a partial or sampled answer. Citations retain edition, license, attribution, and locator information so a caller can trace a result to its source.

The same Go engine is used by multiple interfaces: the MCP server, a command-line application, an HTTP API with a local web UI, a desktop launcher, and the Go package. The MCP interface is therefore one transport over the read-only engine rather than a separate content implementation.

Setup and configuration

The repository does not include the corpus. Before building the database, you must provide the external source roots explicitly; the build command does not use a default path and refuses to run without both roots.

The documented inputs include STEPBible TAGNT and TAHOT data, morphology and lexicon tables, Swete’s Septuagint, Open Scriptures LXX lemma data, public-domain translations, cross-references, and other listed sources. The checked-in sources/sources.json file is the authoritative record for source globs, formats, licenses, and attributions.

The generated SQLite database is disposable. It is not the canonical corpus and can be deleted and regenerated from the supplied files.

Tools and capabilities

The orthotomeo MCP server exposes the engine’s documented operations for:

• Resolving scripture references.
• Retrieving verbatim verses and passages.
• Running lemma and phrase concordances.
• Parsing and lemmatizing text.
• Checking manuscript attestation.
• Rendering citations with source provenance.

These capabilities support questions that require exact textual evidence across multiple editions. They do not provide an interpretation layer, argument generator, or write operation. The server is intended to return the underlying material so a calling model or developer can perform analysis while checking conclusions against the cited results.

Limitations and notes

The orthotomeo MCP server is read-only and does not modify the corpus or database. It also depends on external corpus files that users must obtain and supply under their respective licenses. Results are limited to the editions and source data included in the local build.

Cross-edition alignment is not guaranteed to be one-to-one. Differences in canon, versification, manuscript conventions, and editorial presentation may produce caveats in returned data. The engine favors textual evidence from its loaded sources and does not claim that its output constitutes theological interpretation or a final conclusion about meaning.