What you can ask once connected
Your assistant answers from the published corpus itself — retrieved verses with real citations, not paraphrase from training data. For example:
- “What does the Bhagavad Gita say about acting without attachment to results? Quote the verse.”
- “Show me today's SoulVriti daily verse and explain it in plain words.”
- “Find verses about stillness of mind, then show how the Tao Te Ching and the other traditions echo each other.”
The six tools
Five read-only content tools — the same readers and scorers the product's own search and daily-verse surfaces run on — plus one AI tool with the same grounding, safety checks, and citations as the app:
| Tool | What it does |
|---|---|
list_texts | Every text in the corpus with its counts — the roster, derived live. |
get_verse | One verse by id: original script, transliteration, translations, themes, citation. |
search_verses | Relevance-ranked search across the whole corpus, optionally filtered by tradition. |
daily_verse | The global verse of the day — the same pick every SoulVriti surface shows. |
resonance | Up to three cross-tradition verses that echo a given verse, with shared themes. |
ask_traditions | One question, answered side-by-side by the traditions with verified citations — AI-generated (Cloudflare Workers AI), 2 questions per rolling 24 hours per network. |
Connect in a minute
The server lives at https://api.soulvriti.com/mcp — streamable HTTP, no authentication.
Claude Code
claude mcp add --transport http soulvriti https://api.soulvriti.com/mcpclaude.ai (web & desktop)
Settings → Connectors → Add custom connector → paste https://api.soulvriti.com/mcp (no authentication).
ChatGPT
Settings → Connectors (developer mode) → add a custom MCP connector with the same URL.
Any other MCP client works the same way — point it at the URL above with the streamable-HTTP transport.
Fair use
The connector shares the public content endpoints' abuse limiter: 30 requests per minute per IP (over-limit requests get a 429 with aRetry-After). Plenty for a conversation — just keep agent loops modest and batch questions into fewer, more specific tool calls.
Prefer a reading room to a protocol?
Everything the connector serves is also a quiet, free reading space — or ask one question and hear every tradition answer.