MCP — AI Agent Integration

The Torch exposes an integrated Model Context Protocol (MCP) server alongside its REST API. Any MCP-compatible client — Claude Desktop, Cursor, Windsurf, or a custom agent — can query saints, Scripture, liturgical dates, and sacred sites without writing integration code.

One config block. The entire Tradition, in context.

---

Connecting Your Agent

The MCP server uses Streamable HTTP transport. Add this block to your client’s MCP configuration:

{
  "mcpServers": {
    "torch-and-lily": {
      "type": "http",
      "url": "https://api.torchandlily.com/mcp",
      "headers": {
        "x-api-key": "tl_your_key_here"
      }
    }
  }
}

Replace tl_your_key_here with a real API key from your dashboard.

Note

The MCP server uses the same API key authentication as the REST API. Generate a key once and use it for both.

Client-Specific Setup

Client	Config File Location
Claude Desktop	~/.claude/mcp.json or project .mcp.json
Cursor	.cursor/mcp.json in your project root
Windsurf	.windsurf/mcp.json in your project root
Custom agents	Consult your MCP client library’s documentation

---

Available Tools

The MCP server exposes the same four read engines as the REST API, organized as callable tools. Each tool returns structured JSON.

Veritas — Scripture & Theology

Tool	Description
list_bible_books	List all books of the Bible with metadata
read_bible_chapter	Read all verses in a chapter
read_bible_verse	Read a specific verse with full metadata
read_bible_verse_text	Get plain text of a verse
search_bible	Full-text search across Scripture
list_commentary_authors	List patristic commentary authors
search_commentaries	Search commentaries by book, chapter, verse, or author
list_summa_parts	List the parts of the Summa Theologiae
list_summa_questions	List questions within a part
list_summa_articles	List articles within a question
read_summa_article	Read a full Summa article with objections and replies
search_summa	Full-text search across the Summa
list_opera	List all books in the theological library
search_opera	Search opera by title, author, or keyword
get_opus	Get metadata for a specific book
get_opus_toc	Get the table of contents for a book
read_opus_chapter	Read a chapter from an opus

Devotio — Saints & Prayers

Tool	Description
search_saints	Search saints by name, rank, or patronage
get_saint	Get a saint by slug
get_saint_by_id	Get a saint by database ID
get_saint_feast	Find saints by feast date
list_saint_ranks	List all canonical ranks
list_patronages	List all patronage categories
search_prayers	Search prayers by title or keyword
get_prayer	Get a specific prayer by slug
get_random_prayer	Get a random prayer

Tempus — Liturgical Calendar

Tool	Description
get_daily_ordo	Get today’s liturgical information
get_ordo_calendar	Get the full calendar for a liturgical year
get_movable_feasts	Get movable feast dates for a given year

Corpus — Sacred Site Mapping

Tool	Description
find_nearby_churches	Find churches near a lat/lon coordinate
get_church	Get details for a specific sacred site

API Knowledge Layer

Tool	Description
get_api_schema	Retrieve the full or engine-filtered OpenAPI schema
suggest_endpoint	Find relevant endpoints by keyword or natural language query

Tip

The suggest_endpoint tool is particularly useful for agents building applications on top of the API. Ask it “How do I find prayers about Mary?” and it returns the most relevant endpoints with their parameters and response shapes.

---

Resources

The MCP server also exposes two resources — static data an agent can read without making a tool call:

URI	Description
torch://api/openapi.json	The complete OpenAPI 3.1 specification
torch://api/engines/{engine}	Filtered OpenAPI spec for a single engine (veritas, devotio, tempus, corpus, communio)

---

Example Conversation

With the MCP server connected, an agent can answer queries like:

User: “What saint is celebrated on March 7?”

The agent calls get_saint_feast(month=3, day=7) and returns the result — no code, no HTTP client, no parsing.

User: “Draft a React component that shows today’s liturgical season and color.”

The agent calls get_daily_ordo to understand the response shape, then get_api_schema(engine="tempus") to confirm all available fields, and writes the component with accurate type definitions.

---

Rate Limits

MCP tool calls consume the same rate limit budget as REST API calls: 200 requests per 15 minutes per API key. See Authentication & Rate Limits for details.