Multi-school
Parāśarī, Jaimini, and KP computed in parallel — every result carries school attribution. Schools aren't filters; they're sealed computation contexts with independent CIL passes.
A deterministic Jyotish engine for AI agents — multi-school, 7 languages, every answer traceable to a classical source and a JPL-grade fact.
2,147 keys issued · 1.24M requests served
Every KundaliMCP interpretation is a chain of verifiable facts. The explain tool walks it end-to-end: planetary position → yoga detection → classical citation → interpretive weight. Nothing asserted without provenance.
Life-event prediction: 95.9% recall · 0.00yr period-exact — versus 19.8% recall and 4.33yr period error for the widely-deployed commercial reference. Validated against a 200-fixture corpus and a 14-celebrity panel across Parāśarī, Jaimini, and KP schools.
No wrappers, no adapters. KundaliMCP is a first-class Model Context Protocol server —
tool discovery via tools/list, streaming via SSE,
OAuth 2.1 with PKCE. Any spec-conformant client connects without ceremony.
Claude · ChatGPT · Kimi · any spec-conformant client
Parāśarī, Jaimini, and KP computed in parallel — every result carries school attribution. Schools aren't filters; they're sealed computation contexts with independent CIL passes.
Sanskrit, Hindi, Tamil, Telugu, Kannada, Bengali, English — a structural property of every response, not a localization afterthought. Data-layer multilingualism from day one.
Every rule carries a chapter:verse citation. BPHS, Sāravalī, Phaladīpikā, Jaimini Sūtras — the corpus is the spec. No interpretation without a source.
Full Jyotish computation exposed as MCP tools — discoverable, versioned, documented.
One MCP endpoint. 18 tools. Classical Jyotish — deterministic, cited, multilingual.