KundaliMCP

Jyotish an AI agent can compute, cite and audit.

AI agents are increasingly asked about Vedic astrology. Without a real engine they guess the planets and hide the logic in a prompt. KundaliMCP gives them deterministic computation over the Model Context Protocol, with a classical source attached to every inference.

Business model
API subscription
Status
Live
The KundaliMCP homepage

01 · Where the AI sits

How an AI agent uses KundaliMCP

The agent

Chooses, asks, explains

Any MCP client discovers the tools, picks the right one for the question, and turns the computed result into an answer a person can read. Claude and ChatGPT connect over OAuth 2.1.

KundaliMCP

Computes and cites

Positions, dashas, vargas and yogas are computed the same way every time. Each inference carries the classical text it rests on, from BPHS, Saravali, Phaladeepika, the Jaimini Sutras and others.

Vedaksha

The astronomy

A clean-room Rust ephemeris, validated against NASA JPL Horizons, underneath every number the agent sees.

  • Ten MCP tools

    chat, kundali, kundali_milan, lifemap, muhurat, panchang, pramaan, tithi, get_version and submit_feedback.

  • Classical yogas, qualified

    Each yoga passes a six-stage qualification: detection, eligibility, strength, purity, timing and domain, with cancellation rules applied.

  • Five dasha systems

    Vimshottari plus the Jaimini Chara, Sthira, Niryana Shoola and Maheswara systems, down to the prana level.

  • Sixteen vargas and Shadbala

    The full Shodashavarga set, and six-component Shadbala validated against B. V. Raman's standard horoscope.

  • Safeguards on chat

    A crisis gate stops a message before it reaches a model, and claims are checked against the computed chart.

  • Seven languages

    English, Hindi, Sanskrit, Tamil, Telugu, Kannada and Bengali.

02 · How agents built it

What agents wrote and what people decided

Agents drafted

MCP server scaffolding, JSON-RPC schemas, localisation drafts, candidate source citations.

Humans designed

The yoga taxonomy, isolation between the Parashari, Jaimini and KP schools, the source-attribution model and the interpretive layer.

Gated on every change

Every classical citation is checked against its source text. Every yoga rule and every interpretation is reviewed before it ships.

03 · Stack

What it runs on

A Rust service speaking MCP, so any agent framework that supports the protocol can use it without an SDK.

Protocol
MCP · JSON-RPC 2.0 over Streamable HTTP
Runtime
Rust (axum) on Fly.io
Ephemeris
Vedaksha · VSOP87A and ELP/MPP02
Auth
OAuth 2.1 · API keys
Output
One XML document per call, source-cited
Works with
Claude, ChatGPT, LangChain, CrewAI, AutoGen
Endpoint
mcp.kundalimcp.com/mcp