The Vedic astrology engine your AI can call.
Birthstar MCP is a Model Context Protocol server over the Swiss Ephemeris. Thirty tools for charts, timing, relationships and life patterns — computed, not guessed.
Questions your AI couldn’t answer yesterday.
Not horoscopes. Your actual chart, read against real planetary math — then explained in plain language.
What's changing in my life?
Current dasha, sub-periods and the transits crossing them.
Where is my career going?
Tenth-house indicators, D10 varga, and the periods ahead.
How compatible are we?
Two charts, Ashta-Kuta scoring, and the dynamics underneath.
When should I make the move?
Favourable windows, drawn from timing rather than mood.
Where should I live?
How the same chart reads across places.
Show me my life timeline.
Every major planetary period, birth to now to next.
Understand my child's chart.
Temperament and strengths, without the fortune-telling.
I'm starting a company.
Founder timing, partnership dynamics, launch windows.
The AI interprets. Birthstar calculates.
A language model is very good at meaning and very bad at orbital mechanics. Birthstar takes the half that has to be exact.
- language
- reasoning
- context
- interpretation
- planetary positions to the arcsecond
- 16 vargas
- dasha periods to the day
- transits
- planetary strengths
Ask the same question twice and the math comes back identical. That’s the part you can build on.
30 engines, grouped by what you’d ask them.
Select an engine to see what it computes.
Underneath the conversation is a serious engine.
Chart computation runs a full Swiss Ephemeris integration in about 95 milliseconds. Every subsequent read against that chart handle returns in under a millisecond.
NASA JPL DE431
The ephemeris the engine integrates against.
Swiss Ephemeris
Compiled C, not interpreted approximation.
16 vargas
D1 through D60, exactly divided.
4 dasha systems
Vimshottari, Yogini, Ashtottari, Chara.
Your birth data stays yours.
Date, time and place are held in memory for the length of the computation and then gone. Not logged, not written to disk, not passed to analytics.
Stateless by design
One endpoint. Any client.
{
"mcpServers": {
"birthstar": {
"url": "https://mcp.birthstar.ai/mcp"
}
}
}In Claude Web: Go to Settings → Connectors → Add custom connector, enter https://mcp.birthstar.ai/mcp, and start chatting: “Build my birth chart and describe my rising sign.”
Birth data is processed in memory. Never written to logs, disk, or third-party analytics.
Chart computation runs a full Swiss Ephemeris integration in about 95ms; every cached read after that returns in under a millisecond.
Built on the standard Model Context Protocol — works across Anthropic, OpenAI, Cursor, and custom clients.
What would you build with this?
Ideas, not products — nobody has shipped these yet. The engine is there if you want to.
The engine, as it is running right now.
Every number below is measured, not claimed. Read the same data yourself at /stats.
What happens when your agent calls
A cold chart — new birth data, real ephemeris work — lands around 750 ms. Everything after it reads from the handle in about 0.5 ms. Compute once, then ask as many questions as you like.
Which engines agents actually reach for
Bars are call volume over 30 days; the second figure is median latency. Lime is the one tool that computes; the rest read a chart that already exists. create_chart’s median looks like a read because repeat calls with the same birth data are served from the cache — its cold cost is the p95 above.
Measured on the live server over the last 30 days, busiest day 1,697 calls. Durable totals from the event store, so they survive redeploys — not one process’s counters. 141 calls failed in this window, mostly quota refusals before the demo limit was raised.
Give your AI a birth chart.
One endpoint. Ten seconds. Any MCP client.