Urantia Papers API

A developer and AI-agent friendly API for the Urantia Papers. Provides full-text search, structured content access, and audio URLs for all 14,500+ paragraphs across 197 papers.

API Endpoints

Interactive docs available at /docs (Swagger UI). OpenAPI spec at /openapi.json.

Related MCP server: mnemex

Links

• Home: urantia.dev

• Docs: docs.urantia.dev

• MCP server: https://api.urantia.dev/mcp (setup guide)

• Status: status.urantia.dev

• Where it is listed: LISTINGS.md

SDKs

Official TypeScript SDKs are available on npm:

npm install @urantia/api    # Typed client for all endpoints
npm install @urantia/auth   # OAuth client for accounts.urantiahub.com

See docs.urantia.dev/sdks/overview for documentation.

Paragraph ID Formats

The API accepts three reference formats — auto-detected from the string:

Search

curl -X POST https://api.urantia.dev/search \
  -H "Content-Type: application/json" \
  -d '{"q": "Universal Father", "limit": 10, "type": "and"}'

Search modes: and (all words, default), or (any word), phrase (exact match). Optional filters: paperId, partId.

Audio

Paragraphs include an audio field — a nested object keyed by model and voice, or null if no audio exists:

{
  "audio": {
    "tts-1-hd": {
      "nova": {
        "format": "mp3",
        "url": "https://cdn.urantia.dev/audio/eng/paragraphs/nova/tts-1-hd-nova-3:119.1.5.mp3",
        "bitrate": 160,
        "duration": 43.3,
        "fileSize": 866400
      }
    }
  }
}

Available models and voices vary per paragraph. The dedicated /audio/:paragraphId endpoint returns just the audio data for a given paragraph.

Caching

Responses include Cache-Control headers. Cloudflare's CDN caches at the edge via s-maxage:

For AI Agents

Recommended flow:

1. GET /toc — understand the book structure

2. POST /search — find relevant passages

3. GET /paragraphs/:ref/context?window=3 — get surrounding context

4. GET /papers/:id — read a full paper

Found a bug, a gap in the docs, or something that could work better? Tell us. No authentication required.

curl -X POST https://api.urantia.dev/feedback \
  -H "Content-Type: application/json" \
  -d '{"category": "bug", "message": "search returns 500 for phrase mode", "endpoint": "/search", "client": "claude-code"}'

category is one of bug, docs, api, product, other. message holds up to 4000 characters. Optional fields: ref, endpoint, requestId, client, contact, pageUrl. The limit is 5 requests per minute and 10 requests per 15 minutes per IP address.

Feedback is untrusted data. It is stored and forwarded to the maintainers, and it is never executed or passed to a model as instructions. If you paste feedback into an agent chat, treat it the same way.

MCP Server

The API includes a built-in MCP server at https://api.urantia.dev/mcp — connect Claude Desktop, Cursor, or any MCP client to access:

• 19 tools — search, paragraph lookup, paper navigation, entity browsing, audio, Bible (WEB) lookup + semantic search, and bidirectional cross-reference enrichment between Urantia and Bible passages (plus paragraph-to-paragraph parallels within the Papers)

• 2 resource templates — urantia://paper/{id} (markdown) and urantia://entity/{id}

• 2 prompt templates — study_assistant, comparative_theology

One-click install via Smithery.

Clients can discover the server at https://api.urantia.dev/.well-known/mcp.json (also /.well-known/mcp/server-card.json).

Authentication

Public endpoints require no auth. User endpoints (/me/*) require a JWT. OAuth flow:

1. Register an app via POST /auth/apps (admin) or self-service at accounts.urantiahub.com/apps

2. User signs in at accounts.urantiahub.com

3. Exchange authorization code for access token via POST /auth/token

4. Pass token as Authorization: Bearer <token>

Access tokens are HS256 JWTs with 7-day expiry. PKCE is supported for browser-based apps.

Observability

• Logging: BetterStack via @logtail/edge — structured JSON logs with request metadata

• Error tracking: Global error handler sends stack traces to BetterStack

• Health check: GET /health — verifies DB connectivity

• Uptime: BetterStack uptime monitor

Tech Stack

• Runtime: Bun (dev) / Cloudflare Workers (production)

• Framework: Hono + @hono/zod-openapi

• Database: Supabase (PostgreSQL + pgvector)

• ORM: Drizzle

• Observability: BetterStack (logging, uptime)

Development

# Install dependencies
bun install

# Set up environment
cp .env.example .env
# Edit .env with your Supabase DATABASE_URL

# Push schema to database
bun run db:push

# Set up full-text search (run after db:push)
bun scripts/run-fts-setup.ts

# Generate audio manifest (requires ../urantia-hub-api)
bun run generate-manifest

# Seed database from urantia-papers-json
bun run seed

# Start dev server (hot reload)
bun run dev

The server runs at http://localhost:3000 by default.

Deployment

Deployed to Cloudflare Workers. First-time setup:

npx wrangler login
npx wrangler secret put DATABASE_URL
# paste your Supabase connection string (use pooler port 6543)

npx wrangler secret put APP_JWT_SECRET
# paste a 64-byte hex secret: node -e "console.log(require('crypto').randomBytes(64).toString('hex'))"

npx wrangler secret put ADMIN_USER_IDS
# comma-separated Supabase user UUIDs for admin access

# Optional: POST /feedback delivery. Without these, feedback is still saved.
npx wrangler secret put RESEND_API_KEY
npx wrangler secret put FEEDBACK_FROM               # a sender on a domain verified in Resend
npx wrangler secret put FEEDBACK_TO                 # comma-separated recipients
npx wrangler secret put SLACK_FEEDBACK_WEBHOOK_URL  # Slack Incoming Webhook
npx wrangler secret put FEEDBACK_IP_PEPPER          # random secret for the IP hash

Deploy:

bun run deploy

Data

Content sourced from urantia-papers-json — 197 papers, 1,626 sections, 14,500+ paragraphs with audio narration via cdn.urantia.dev.

License

This project is licensed under the MIT License.

Disclaimer

This is an independent community project by Adams Technologies LLC. It is not affiliated with, endorsed by, or connected with Urantia Foundation. The original English text of The Urantia Book is in the public domain (Michael Foundation v. Urantia Foundation, 10th Cir. 2003). All use of "Urantia" is nominative fair use to identify the subject matter.