Back to home

Vedic astrology API · ChatGPT Action · MCP for Claude & Cursor

Put a real Vedic astrology engine inside your AI

Eternal Evals is a computation-first Jyotish engine on the Swiss Ephemeris. Most astrology sites publish articles about charts. We let your AI compute one — deterministically, for any exact birth — then read it section by section. Connect it four ways.

1 · ChatGPT

A ready-made Custom GPT that calls the engine through a GPT Action. Zero setup — just chat.

Open the GPT →

2 · Claude & Cursor (MCP)

An MCP server exposing compute_chart, read_section and list_sections as native tools.

Set up MCP →

3 · REST API

A plain JSON API. Compute a chart, then read any of ~70 sections on demand.

Read the API →

4 · In-browser agents (WebMCP)

The agent inside the visitor's browser (Gemini in Chrome, agentic browsers) calls our tools directly — no install.

About WebMCP →

What the engine computes

Every result is a real, deterministic computation — not content keyed to a Sun sign. From an exact birth date, time and place, the engine (Swiss Ephemeris DE441, Lahiri ayanamsa, whole-sign houses by default) computes the sidereal chart and its full classical stack: planets, houses and lagna; 16 divisional (varga) charts; Vimshottari plus 21 conditional dasha systems; shadbala strength in rupas; ashtakavarga bindus; 100+ yogas; nakshatra structure; and live transits. Identical input yields byte-identical output, and every number is auditable. The engine only computes — your model does the interpretation.

How this differs from other astrology APIs

Several Vedic astrology APIs and MCP servers now exist. What sets this one apart is not the number of endpoints — it is what an AI agent can safely do with the output:

The REST API

Two steps: compute a chart to get a chart_id and the list of section names, then read the sections you need (the full chart is large, so you fetch sections on demand).

# 1. Compute a chart
curl -s -X POST https://eternalevals.com/api/v1/chart \
  -H "Content-Type: application/json" \
  -d '{"date":"1990-07-07","time":"05:00","place":"Panipat, India"}'
# -> { "chart_id": "chart-abc123", "section_count": 71, "sections": [ ... ] }

# 2. Read a section (dot-paths work too: houses.10, planets.Saturn)
curl -s https://eternalevals.com/api/v1/chart/chart-abc123/section/vimshottari_dasha

Free with a per-IP rate limit. Add -H "Authorization: Bearer <key>" for a higher limit. The full machine-readable contract is at /openapi.json; a map of the whole site for AI crawlers is at /llms.txt.

MCP server for Claude & Cursor

The Model Context Protocol (MCP) lets assistants like Claude Desktop, Cursor and Cline call external tools. Our MCP server wraps the API as three tools — compute_chart, read_section, list_sections — so the assistant computes and reads a real Vedic chart natively, no copy-paste.

# Get the server + its dependencies
curl -O https://eternalevals.com/mcp/eternalevals_mcp.py
pip install mcp httpx

Then point your client at it. For Claude Desktop, add this to claude_desktop_config.json (Settings → Developer → Edit Config):

{
  "mcpServers": {
    "eternal-evals": {
      "command": "python",
      "args": ["/absolute/path/to/eternalevals_mcp.py"],
      "env": { "ETERNALEVALS_API_KEY": "" }
    }
  }
}

Cursor and Cline use the same command and env in their MCP settings. The API key is optional — leave it blank to use the free anonymous limit, or set a bearer key for a higher one. Restart the client and you'll see the three tools appear.

WebMCP — for the agent inside the browser

The MCP server above is for agents outside the browser. WebMCP is the other side: a browser API (navigator.modelContext) that lets our pages hand tools to the AI agent running inside the visitor's browser — Gemini in Chrome, agentic browsers, or an agent extension. There's nothing to install; when a WebMCP-capable agent is on an Eternal Evals tools or panchang page, the tools are simply available, and they run with the visitor's own session (so a signed-in user's agent can read their own charts, no key).

On any tools page the agent sees the full catalogue — compute_vedic_chart, read_chart_section, and a read_… tool for every calculator; the panchang page adds get_panchanga and find_muhurta. Each is a thin, deterministic wrapper over the same engine, so the agent answers from computed numbers rather than scraping the page.

WebMCP is a W3C Community Group proposal, live in a Chrome origin trial (Chrome 149+) and natively in Edge 147+. Full walkthrough: Let your browser's AI agent compute your Vedic chart.

Why compute, not content

Ask a generic astrology chatbot about your Saturn and it may invent a placement that was never calculated — confident prose over unstable facts. Eternal Evals separates the two jobs: a deterministic engine computes the chart first, and the AI reads what the engine produced. That is the difference between a tool that talks about astrology and one that answers about your chart — and it is why an AI grounded in this engine can show the exact placement, strength or period behind any claim.

Frequently asked

Is there a Vedic astrology API?

Yes — a free JSON REST API at /api/v1. POST birth details to /api/v1/chart for a chart_id and the list of ~70 sections, then GET the sections you need.

How do I use the engine in ChatGPT?

Open the Eternal Evals Custom GPT, or build your own GPT by importing https://eternalevals.com/openapi.json as an Action. Give it a birth date, time and place and it computes a real chart, then interprets it.

Does it work with Claude or Cursor?

Yes, via MCP. Add the MCP server above to Claude Desktop, Cursor or Cline and the assistant gets compute_chart, read_section and list_sections as native tools.

Is it free?

Yes, with a per-IP rate limit. A bearer API key raises the limit for higher-volume use.

What ayanamsa and house system does it use?

Lahiri (Chitrapaksha) ayanamsa and whole-sign houses. The engine implements 12 ayanamsa systems and 8 house systems internally, but every chart the API computes uses Lahiri and whole-sign, and says so in its convention manifest.