Skip to content

Vedic astrology computation
reimagined from first principles.

KundaliMCP is the first agentic-AI-native Jyotish computation platform — a Model Context Protocol toolkit that gives AI agents the ability to cast, analyze, and interpret kundali charts with full deterministic transparency. Not a chatbot wrapper. Not a lookup table. A computational engine that models the full complexity of Jyotish as a graph problem.

Get an API keyRead the docs
95.9%
Life-event recall
0.39′
Mean planetary error
18
MCP tools
7
Languages
3
Schools
Zero
PII ever stored

A computation engine, not a chatbot.

Existing Jyotish software treats chart computation as a solved problem and interpretation as a black box. KundaliMCP treats both as engineering problems. Computation is a graph traversal, not a lookup table. Interpretation is deterministic rule application, not LLM hallucination.

Every conclusion has a provenance chain back to the astronomical facts. Every rule carries its school attribution — no silent bias. The same input always produces the same output from the deterministic layer; the interpretive layer is tunable but never touches the astronomical foundation.

The technologies of 2026 — graph databases, defeasible reasoning engines, MCP protocol, multi-pass analysis on commodity hardware — make it feasible to model Jyotish the way it was designed to work: as a recursive, multi-dimensional system where every factor affects every other factor.

Agents call compute_chart, explain, forecast_events — and get back fully traceable, multi-school, bilingual results grounded in classical texts.

Why KundaliMCP is different.

01

Agentic-AI-First Architecture

Not a REST API retrofitted for AI agents. Built from the ground up as an MCP server. Agents discover tools via tools/list, call them with structured JSON, and receive structured results. Spec-conformant OAuth 2.1 + RFC 9728 means one-click connect from Claude, ChatGPT, Kimi, or any MCP client shipping tomorrow.

MCP Streamable HTTPJSON SchemaOAuth 2.1SSE streaming
02

Deterministic Core + Calibratable Interpretation

Hard separation between IDL (astronomical facts, Shadbala, yoga detection — always identical for the same input) and CIL (weights, thresholds, interpretations — tunable per school). You can trust the facts absolutely and improve interpretations incrementally without touching the computational foundation.

VSOP87AELP/MPP02IDL/CIL separationSchool-parameterized
03

Multi-School Without Bias

Pārāśara, Jaimini, KP — each is a sealed computation context. A SchoolProfile is a first-class parameter on every call. When two schools disagree, both are presented with attribution. The system never picks a winner silently.

Pārāśara (BPHS)Jaimini SūtrasKP PaddhatiNāḍī norms
04

Full Provenance Chains

Every interpretation can be traced back to its root through a complete 6-stage chain: IDL astronomical fact → yoga detection → Vivek qualification → CIL weight → school attribution → text. The explain tool returns this full chain for any claim. No black boxes.

6-stage Vivek pipelinePer-rule weightsClassical citationsBPHS chapter:verse
05

Graph-Based Multi-Pass Analysis

Planetary relationships modelled as a petgraph graph — planets, houses, signs, nakshatras as nodes; aspects, rulerships, dispositor chains as edges. Influence propagates recursively until convergence. Not a single-pass rule table.

Anvaya (petgraph)Dispositor chainsRecursive influenceYoga activation
06

Privacy by Design

Birth data is never stored. Not in logs, not in error messages, not in any persistent store. Stateless computation: data enters via MCP call, runs in memory, results exit. Ephemeral cache is opt-in, caller-keyed, AES-256-GCM encrypted.

StatelessNo PII storedAES-256-GCM cacheZero log exposure

Ephemeris to product. All in-house.

No third-party ephemeris. No vendor lock. The entire stack from raw sky coordinates to MCP response is owned and auditable. A new primitive flows from sky to screen in days.

Ephemeris
Vedaksha
BSL 1.1

Clean-room pure-Rust · VSOP87A + ELP/MPP02 · BSL 1.1

Compute core
KundaliMCP Engine
Live

Single statically-linked Rust binary · axum · MCP handler · auth · billing

Graph layer
Chart graph

petgraph StableGraph · per-school sealed context · dispositor chains

Reasoning layer
Defeasible engine

Rust-native · 6-stage yoga qualification · cancellation rules · provenance

Localization
Ontology store

7-language · JSON-LD · en hi sa ta te kn bn · compile-time baked

Edge gateway
Cloudflare Workers

wasm32 deploy · KV cache · R2 observation store · Queues

Every number is reproducible.

The audit harness at audit/life_events/ validates against BPHS, Sāravalī, Phaladīpikā, Jaimini Sūtras, and JPL DE421. Run it yourself: cargo run --release --bin audit-life-events

95.9%
Life-event recall
vs 200-fixture classical corpus · reference baseline: 19.8%
0.00yr
Period-exact timing
Vimśottarī dasha corpus · reference baseline: 4.33yr avg
0.39′
Mean planetary error
vs JPL DE421 · 343 positions · max 0.80′ · zero sign mismatches

Every rule cited. No silent assumptions.

Every life-event trigger, yoga definition, and dasha rule in the engine is derived from a named classical source at chapter:verse level — not from a prior-generation reference and not from convention. If a source contradicts another, both are recorded with their respective school attributions.

BPHS
Bṛhat Parāśara Horā Śāstra
Pārāśara
Sāravalī
Kalyāṇavarman
Pārāśara
Phaladīpikā
Mantrēśvara
Pārāśara
Jaimini Sūtras
Mahārṣi Jaimini
Jaimini
Uttara Kālamṛta
Kālidāsa
Pārāśara
JPL DE421
NASA / USNO
Ephemeris

Built for builders.

AI agent builders

Add classical Jyotish computation to any MCP-compatible agent — Claude, ChatGPT, LangChain, CrewAI, AutoGen, or your own. The protocol is the product; no adapter layer, no SDK, no glue code.

Jyotish software developers

Use KundaliMCP as a computation backend. 18 tools, all five dasha systems, all 16 divisional charts, 7 languages — production-ready today.

Researchers

Study traditional knowledge systems with modern computational tools. Every rule is cited; every result is reproducible; the audit harness is public.

Practitioners

Transparent, multi-school analysis. See the provenance chain behind every interpretation. No black boxes. No silent school blending.

ArthIQ Labs

Human intent, meets AI-native execution.

ArthIQ Labs is an AI-native incubation studio, built on agentic engineering. Coding agents handle first-pass execution across the software lifecycle; humans own architecture, tradeoffs, and outcomes.

The work of a quarter, in a week. Live products at roughly a tenth the cost of a traditional team. No concept decks, no endless pilots — concept to launch in 7–21 days.

Visit arthiq.net →
Advisory & Consulting

Strategy, architecture, and roadmapping for AI-native businesses.

Build Services

End-to-end design and shipping of AI-native products. Concept to launch in 7–21 days.

Managed Services

Operations, support, and agentic process automation for AI systems in production.

Generalinfo@arthiq.netX@ArthIQLabs

Ready to add Jyotish to your agent stack?

Get a key in 60 seconds. No credit card required on the free tier. 18 tools, 7 languages, 3 schools — live at mcp.kundalimcp.com/mcp.

Get API Key Read the docs
॥ श्री गणेशाय नमः ॥