An AI agent that does full Jyotish chart analysis, not vibes

What it does

This is a toolkit of eight specialized skills for any client that reads the Claude Skill format (Claude Code, Codex, and others). You give it birth date, time, and place; vedic-calculator computes a full Vedic chart — planets, houses, Nakshatras, 15 divisional charts, Shadbala, and a complete Vimshottari MD/AD/PD timeline — and downstream skills consume a single validated structured_data.md for natal, career, relationship, birth-time rectification, and synastry analysis. There’s also a standalone Prashna (horary) skill that charts the moment a question is asked, deliberately kept separate from the natal pipeline.

The interesting bit

The engineering discipline is unusual for astrology tooling: a 16-point validation system checks things like Rahu-Ketu opposition, planetary war, Ayanamsa, and Dasha continuity before any analysis runs, and it explicitly refuses to equate “birth certificate exists” with “all divisional charts are reliable.” The calculator generates all 729 Pratyantar Dasha rows up front but serves them through dasha_query.py on demand, so the model context isn’t drowned in timeline tables.

Key highlights

• Charts computed from raw birth info — no dependency on external sites or pre-made PDFs (though JHora exports can be cross-checked)
• Birth-time rectification scans all candidate times against five-plus life events rather than confirming a favorite
• Synastry deliberately withholds a “compatibility percentage” and won’t reverse-engineer the relationship type from the charts
• Trilingual output: Chinese, English, and Japanese for conversation, reports, and Q&A
• AGPL-3.0 licensed, Python 3.8–3.13

Caveats

• The README recommends installing all eight skills; there’s a Codex-specific patch section, suggesting platform quirks exist
• The public repo ships the “standard edition” of vedic-core — implying a fuller version exists elsewhere
• Prashna rules are filtered through KN Rao/Bharatiya Vidya Bhavan practice compatibility, so interpretive authority is a specific lineage, not a neutral canon

Verdict

Worth a look if you build agent tooling and want to see what rigorous data-contract design looks like in an unlikely domain — the validation and context-management patterns are genuinely transferable. Skip it if you wanted a quick horoscope API; this is a deep, opinionated pipeline.