# orthotomeo > A read-only scripture-study engine over a cited biblical corpus. Answers are > retrieved from published source files, not recalled - and a query returns > every match or an explicit error, never a sample. The corpus is a derived SQLite database built from openly licensed published editions: the Greek New Testament and Hebrew Old Testament with morphology and Strong's tagging, two Septuagint witnesses, Strong's lexicons, and several public-domain English translations. Every row carries the edition it came from, its locator inside that source, and its licence. ## What it is useful for - Concordance work. Every occurrence of a lemma or Strong's number across the tagged corpus, complete or refused with an exact count. - Morphology and lexicon lookups for a specific word in a specific verse. - Reading one reference across several editions at once, including where a reference does not exist in an edition - absence is reported as data. - Manuscript attestation, reported as neutral data rather than argued. ## Connect over MCP https://orthotomeo.app/mcp Twelve tools over the same engine as the HTTP API. Full reference below. ## HTTP API Every endpoint is GET and returns JSON. There is no mutation surface. - https://orthotomeo.app/verse - one reference, one citation per edition - https://orthotomeo.app/passage - a contiguous single-book range - https://orthotomeo.app/concord - complete-or-fail concordance - https://orthotomeo.app/count - how many a query would match, before running it - https://orthotomeo.app/interlinear - a verse word by word - https://orthotomeo.app/define - lexicon entry for a Strong's number - https://orthotomeo.app/parse - morphology for one word - https://orthotomeo.app/attest - manuscript attestation for one word ## Reference - Full API and MCP documentation: https://orthotomeo.app/docs - Web query tool: https://orthotomeo.app/app/ - Source: https://github.com/jrainsberger/orthotomeo ## Limits The public instance is rate limited per day. A query matching more than the result cap is refused outright with the exact count rather than truncated - completeness is the point, so a partial answer is treated as a failure.