Skip to main content
The Urantia Papers API has a built-in Model Context Protocol (MCP) server. An AI agent that connects to it can read and search the Urantia Papers and quote them by reference.
It has 28 read-only tools, 2 resource templates, and 2 prompts. It needs no key, no account, and no local process. Not a developer? See AI Assistants for steps in Claude, ChatGPT, Grok, and Gemini.

Set up

Claude Code

Install the plugin. It adds the server and a research skill:
Or add the server only:

Cursor

Install the plugin from cursor.directory/plugins/urantia-papers, or add the server to your MCP settings:

Gemini CLI

Install the extension. It adds the server and the research skill:

ChatGPT

Install the Urantia Papers plugin from the ChatGPT plugin directory.

Claude, Grok, and Gemini

Add the address above as a custom connector. AI Assistants has the steps for each one.

Any other client

Use the same mcpServers block as for Cursor. The server works with any client that supports the Streamable HTTP transport.

Where it is listed

Reference

Available Tools

The MCP server exposes 28 tools using dot-notation names organized into a navigable namespace tree. All tools advertise read-only annotations and ship with output schemas for structured responses.

Structure & Navigation

Paragraphs

Entities

Audio

Bible (World English Bible)

Works (the Bible and 21 other texts)

Quotes and citations

Editions

Cross-reference enrichment on existing tools

paragraphs.get, paragraphs.random, search.fulltext, and search.semantic accept three optional booleans that attach pre-computed semantic neighbors to each result: All three can be combined. Parallels are pre-computed via text-embedding-3-large cosine similarity, so adding them is cheap. They are found by meaning, so they link passages on the same subject also when the words differ. See How the links are made.

Spanish text

toc.get, papers.list, papers.get, papers.sections, paragraphs.get, paragraphs.context, and paragraphs.random accept lang: "es" for the Spanish translation of the text and titles. Each Spanish paragraph has language and a translation label (an AI translation, CC0). search.fulltext and search.semantic also accept lang: "es": the keyword search reads the Spanish text, and the semantic search takes a question in Spanish and returns the Spanish paragraphs. See Translations.

Editions of the text

paragraphs.get, paragraphs.context, and papers.get accept edition for an older edition of the English text: "1955", "srt-2009", "srt-2015", "srt-2018", "srt-2021", or "srt-2024" (the default). include_variants: true adds each change to a paragraph since 1955, with its year, 1955 page, and reason. editions.list and editions.variants list the editions and all the changes. See Editions of the text. The answer to initialize has instructions: a short guide to the references, the tools, and the editions, for the agent to read.

Resources

In addition to tools, the server exposes two resource templates clients can read directly:

Prompts

The server also publishes two reusable prompt templates for common study workflows:

Example Prompts

Once connected, try asking your AI agent:
  • “Search the Urantia Papers for passages about love”: uses search.fulltext
  • “What do the Urantia Papers say about what happens after death?”: uses search.semantic
  • “Read Paper 1 about the Universal Father”: uses the urantia://paper/1 resource
  • “Show me paragraph 2:5.10 with surrounding context”: uses paragraphs.context
  • “Find all entities of type ‘place’”: uses entities.list
  • “What entities are mentioned in paragraph 0:0.1?”: uses paragraphs.get with include_entities
  • “Compare what the Urantia Papers teach about the soul with Stoic philosophy”: uses the comparative_theology prompt
  • “Find Urantia paragraphs related to Matthew 5:3”: uses bible.verse.urantia_parallels
  • “Search the Bible for passages about forgiveness and show the related Urantia teachings”: uses bible.search.semantic
  • “Show me 0:0.1 with both Bible parallels and related Urantia paragraphs”: uses paragraphs.get with include_bible_parallels and include_urantia_parallels

How It Works

The MCP server uses Streamable HTTP transport, which means:
  • No local process: it runs on the same Cloudflare Worker as the API
  • Stateless: each request creates a fresh server instance (no sessions to manage)
  • Same rate limits as the REST API (200 requests per minute for each IP address)
  • Same data: MCP tools query the database directly, returning the same results as the REST endpoints

MCP vs REST

Both give you access to the same data. Choose based on your use case:

Docs MCP Server

Mintlify provides a hosted MCP server that lets AI agents search these documentation pages, useful for discovering endpoints, understanding usage patterns, and learning the API.

Setup

Add the address to your client as a second server:

Available Tool

Example Prompts

  • “How do I search the Urantia Papers API?”
  • “What paragraph reference formats does the API support?”
  • “Show me how to use the entities endpoint”
  • “What audio voices are available?”

Building a Custom MCP Server

If you need custom logic (e.g., combining multiple tools, caching, or preprocessing results), you can build your own MCP server that calls our REST API. See the Build an MCP Server tutorial for a step-by-step guide.