Launching AstroWay API: Comprehensive Astrology Endpoints with Swiss Ephemeris Accuracy

We’re announcing AstroWay API today — a calculation backend for developers building astrology apps, horoscope products, dating-compatibility features, and AI agents that need to reason about charts.

This post covers what’s actually in the API, the unusual choices we made, what’s deliberately not in scope, and how we got here. No fluff.

🕓 Recorded 2026-04-22. Some numbers and claims here are stale. Current state: 705 endpoints (vs 87 at launch), 577 CI snapshot tests (vs 183), MCP server @astroway/mcp, TypeScript @astroway/sdk, Python astroway, and PHP astroway/sdk all shipped (published 2026-05-09 on npm, PyPI, Packagist). Up-to-date status — see /faq and changelog.

A REST API on top of Swiss Ephemeris (LGPL) compiled to WASM and hosted as a Node.js service — now with AI interpretations on top (GPT-4 / Claude / Llama 3.3 / Gemini / DeepSeek / Mistral failover chain) layered over the real calculations. Same engine professional astrologers have used in Solar Fire, Kepler, Astro Gold and Janus for 30+ years — sub-arcsecond precision, validated against the NASA JPL ephemeris and Astrodienst’s official swetest reference (Astro.com).

Numbers:

• 705 endpoints covering Western natal, synastry, transits, progressions, directions, returns, harmonics, fixed stars, Arabic parts, astrocartography, plus a complete Human Design suite
• 183 snapshot tests in CI against real swetest CGI output and the NASA Eclipse Catalog — every PR runs them
• p50 < 100ms, p95 < 500ms for typical calculation endpoints
• MCP server: @astroway/mcp — exposes the API as tools for Claude, GPT, Cursor and other LLM agents
• Official SDKs shipped 2026-05-09: TypeScript @astroway/sdk, Python astroway, PHP astroway/sdk — auto-generated from OpenAPI 3.1, retry + idempotency built in
• Credit-based pricing: Free tier (10,000 credits/month, no card), paid plans from $5/month

Full reference: api.astroway.info/docs/api/. Live status: api.astroway.info/status.

Three years ago we started building consumer astrology apps. The calculation layer was the hardest part — far harder than the UI or the content. Existing services either:

1. Had shallow coverage (10–30 endpoints, mostly natal-only)
2. Were expensive at scale ($0.02–$0.10 per request without bulk discounts)
3. Used proprietary engines with accuracy you couldn’t audit
4. Vedic-first, with Western treated as an afterthought

So we vendored Swiss Ephemeris ourselves and wrote our own calculation layer — first for the app.astroway.info consumer product, later for the astroway.info Marketing site that powers daily horoscopes for hundreds of thousands of monthly visitors.

After three years of pushing this code in production we’ve come to two conclusions:

1. The same calculation engine should be available to other developers — not because they can’t write it themselves, but because the library + WASM + caching + accuracy CI stack is six months of work nobody should redo.
2. Our Human Design coverage (12 endpoints — chart, type, authority, profile, gates, channels, circuits, planetary activations, incarnation cross, definition) is genuinely unique. No competitor offers the full HD stack via API.

That’s the launch.

We have /nakshatras, /ashtakavarga, and divisional-chart endpoints, but no Dasha, no Panchang, no Ashtakoota, no KP. The reason: Vedic has unresolved disagreements between traditions (ayanamsa wars, KP cusp disputes, dasha-school choices). Half-baked Vedic competing against Prokerala and VedicAstroAPI would just get criticised.

Decision locked: Vedic suite is a separate Phase 3, after we have traction on Western and time to validate against Jagannatha Hora and Parashara’s Light. Until then we’d rather not ship it than ship it badly.

A single /chart call costs about 50× less compute than a 365-day /forecast-calendar. Flat per-request pricing would either price out light users or subsidise heavy ones. Credits map cost to actual work — see the pricing page for the per-endpoint table.

Free-tier responses include a _footer field:

{
  "data": { ... },
  "_footer": "Powered by astroway.api — get your own key at api.astroway.info"
}

It costs us nothing and gives free users a small reason to upgrade. Removable on Hobby+ via key config. We borrowed the pattern from Zapier and RevenueCat.

/rectification (solving birth time from life events using a likelihood-grid search) is computationally expensive — a single call takes 10–60 seconds and tens of thousands of intermediate Swiss Ephemeris evaluations. We capped it at the Pro plan ($59/month) and credit-priced it heavily because abuse would tank shared infrastructure for everyone else.

/interpret/* endpoints are routed through our private AI gateway sitting on top of the best-of-breed market LLMs (Google Gemini → Groq → OpenRouter → Cerebras → SambaNova → Mistral) with automatic failover. We don’t lock you into our prompts — every interpretation response includes the raw calculation data, so you can re-prompt with your own LLM if ours doesn’t fit your voice. We log full prompts + outputs to an audit table for B2B compliance (90-day retention).

• Vedic suite — Phase 3, post first traction (see above)
• Webhooks for usage thresholds — implemented but UI is rough; iterating
• Public widgets (<script src="api.astroway.info/widget/natal.js">) — Phase 7
• Plugin for the main Marketing site — separate project, planned for the astroway.info ecosystem

1. Read the docs first if you want the why: api.astroway.info/overview
2. 5-minute quick start with code: Astrology API Quick Start
3. Pricing including the per-endpoint credit table: api.astroway.info/pricing
4. Sign up — free plan, no card: api.astroway.info/dashboard/sign-up

If you build something with the API — or hit a bug, or want an endpoint we don’t have yet — reach out. We’d rather hear from a hundred developers asking for things than launch features nobody uses.

Questions, requests, weird edge cases: hello@astroway.info.

Roadmap is public in PRODUCT-PLAN.md — what’s done, what’s coming, what’s deferred and why.

Welcome.