Specialised AI tools for ecclesiastical Latin
Extract dates, places, names, scripture allusions, and catalogue-ready metadata from Catholic Latin records — necrologies, chapter acts, episcopal letters, charters. Connects to Claude, Codex, Cursor, Windsurf, and any MCP-compatible AI assistant.
Church Latin AI connects to your existing AI assistant. It is not a standalone web editor — you need Claude, Codex, Cursor, or a similar tool to use it.
A single necrology entry, processed in seconds.
Latin input
Extracted metadata
| Name | Joannes Brady |
| Order | Discalced Carmelites (O.C.D.) |
| Born | 1845-08-30 (confidence: high) |
| Professed | 1868 |
| Ordained | 1872 |
| Office | Prior, Dublin Convent (5 years) |
| Place of death | Dublin, Ireland |
| Date of death | 1920-03-15 (confidence: high) |
| Age at death | 75 ✓ cross-checked |
| Years professed | 52 ✓ cross-checked |
Ready to export as AtoM CSV, Dublin Core, EAD XML, or JSON.
Ecclesiastical Latin has its own conventions — Kalends and Ides for dates, diocese adjectivals for places, clerical abbreviations that no general-purpose AI reliably handles. These tools are built around that domain knowledge.
Kalends, Nones, Ides. Feast-day anchors ("in festo S. Michaelis"). Pontifical regnal years. Indictions. All converted to ISO 8601.
2,300+ Latin diocese adjectivals mapped to modern cities and countries — Ireland, Italy, Germany, Africa, the Americas, Asia. Handles declined inflected forms.
37,228 Clementine Vulgate verses in full-text search. Detects verbatim quotations, partial citations, and allusions. Returns the Douay-Rheims translation.
Extracted metadata exports to AtoM CSV (ISAD(G)), Dublin Core, EAD 2002 XML, or plain JSON — ready to import directly into your catalogue.
50 tool calls per month. No credit card required.
Enter your email and your API key arrives in seconds.
Pay by card through Stripe. Your API key is emailed to you automatically. Cancel any time.
Payments handled securely by Stripe. API key sent to your billing email within minutes of subscribing.
Try free (50 calls/month, no card needed) or subscribe above for higher limits. Your key arrives by email.
Use the MCP endpoint URL and your API key in your client's configuration.
Paste a Latin text and ask your AI assistant to extract metadata, find toponyms, or detect a scripture allusion.
Run once in your terminal:
Run once in your terminal (the key is read from the environment variable at startup):
Then add your key to your shell profile: export CHURCH_LATIN_KEY="YOUR_API_KEY"
Paste this into your MCP config file and replace YOUR_API_KEY:
| Claude Desktop | ~/Library/Application Support/Claude/claude_desktop_config.json (Mac) / %APPDATA%\Claude\claude_desktop_config.json (Windows) |
| Cursor | .cursor/mcp.json in your project folder |
| Windsurf | ~/.codeium/windsurf/mcp_config.json |
The server accepts your API key as either an X-Api-Key header or an Authorization: Bearer token — use whichever your client supports.
Once connected, paste this into your AI assistant to confirm everything is working:
You should see the AI call expand_clerical_titles and convert_ecclesiastical_date in sequence, then give you a structured result. If you see a 401 error, your API key is missing from the config. If the server is unreachable, restart your AI client after adding the config.
The tools are optimised for Catholic ecclesiastical Latin — necrologies, chapter acts, episcopal letters, professions of faith, charters, and similar records from religious orders and diocesan archives. They work best with 16th–20th century Latin, though date and place tools handle medieval documents too.
Any client that supports the Model Context Protocol (MCP) with HTTP transport: Claude Code (CLI), Claude Desktop, Codex CLI (OpenAI), Cursor, Windsurf, and others. The server follows the MCP standard and requires no special configuration beyond the endpoint URL and your API key.
No document text is stored. The server processes each request in memory and returns a result. The only data retained is your account record (email, plan, and API key) for billing and authentication.
Yes — manage your subscription at any time through the Stripe customer portal. Cancellation takes effect at the end of the current billing period.
Both. Your AI assistant handles translation using its own language capabilities — Church Latin AI adds specialised tools on top: accurate date conversion, toponym resolution, abbreviation expansion, scripture identification, and structured metadata extraction. Think of it as giving your AI assistant a set of expert reference tools it can call on demand.
You need a working installation of one supported AI assistant (Claude Code, Claude Desktop, Codex CLI, Cursor, or Windsurf) and your Church Latin AI API key. Once configured, you simply describe what you want in plain English — the AI decides which tools to call. No programming required.
The free tier gives you 50 tool calls per month with no credit card required. You will need at least one MCP-compatible AI client installed — Claude Desktop is the easiest starting point for non-technical users. Download Claude Desktop here.
Document text is never stored on this server — each request is processed in memory and discarded. However, text you paste into your AI assistant is processed by that assistant's provider (Anthropic for Claude, OpenAI for Codex, etc.) under their own privacy policies. For restricted or confidential records, check your AI provider's data handling terms before use. A self-hosted option for institutions with strict data governance requirements is on the roadmap.
All extraction tools include a requires_human_review flag and confidence ratings. Date conversions are cross-checked against age and profession-year formulas where possible. The tools are designed as a first-pass accelerator, not a replacement for archival judgement — every result should be verified before entering a catalogue.
Church Latin AI is built by Boyne Archives, an Irish archival services company with active contracts cataloguing the records of Catholic religious orders and dioceses. The tools are used in live archival work — not built speculatively. The ecclesiastical gazetteer, abbreviation lists, and date conversion rules are drawn from primary archival sources and tested against real documents.
Reasonably well. The normalisation tool corrects common OCR errors (u/v, i/j, ligatures) before other tools run. Abbreviation expansion and toponym lookup work on partial or inconsistently spelled text. Very poor quality OCR — garbled words, broken line endings, missing characters — will reduce accuracy and trigger human-review flags.
AtoM CSV (44-column ISAD(G) format), Dublin Core, EAD 2002 XML, and JSON are all supported export formats on the Professional and Team plans. The CSV can be imported directly into AtoM. For ArchivesSpace, use the JSON output and map fields to your import template. Excel can open any CSV export directly.
Each tool invocation counts as one call, regardless of how many documents it processes. Sending 50 documents to the batch triage tool uses 1 call from your monthly allowance, not 50.
Tool calls return a 401 response and your AI assistant will report that the server rejected the request. Your limit resets on the first of each month. You can check your current usage at any time: https://churchlatin.ai/usage?key=YOUR_API_KEY
Contact info@archiveshosting.com to discuss annual contracts, pilot projects, or custom volume for larger institutions.