Astro Agents vs Swiss Ephemeris vs computing it yourself

An honest comparison for financial astrology and backtesting: where a hosted, verifiable API saves you work, and where a local library is simply the better tool.

Astro Agents APISwiss EphemerisDIY (Skyfield / SPICE + JPL kernels)
What it isA hosted HTTP API (REST, MCP, A2A) and a Python clientA C library (with bindings) you embedYour own code on Skyfield or NASA's SPICE, with JPL kernels
EphemerisNASA/JPL DE440 (1800-2200 served)Its own compressed files derived from JPL ephemerides; can also read JPL filesAny JPL kernel you download (DE440, DE441...)
AccuracyWithin 0.024″ of JPL Horizons on 12,400 values (published benchmark)Very high: Astrodienst documents agreement with JPL far below an arcsecondAs good as the kernel and your frame handling: the same class
Event calendarReady: ingresses, stations, lunations, eclipses with type, aspects, parallels, out-of-bounds, to the minuteBuilding blocks (eclipse and crossing functions); the scan and bisection are yoursYours to write: root finding for every event kind, eclipse classification
Gann lines, Mean of 5, Circle of 8, price levelsBuilt inYours to writeYours to write
Time zones and historical DSTResolved for you from the placeYours (the library takes UT/TT)Yours
Proof of what you usedSHA-256 of every input and result; free /v1/verifyNone built inNone built in
CostPer call (from $0.01) or prepaid packs; nothing to installFree under AGPL-3.0, or a paid professional licence from Astrodienst for closed-source useFree (Skyfield is MIT; JPL kernels are public)
Offline and volumeNeeds the network; rate-limited; paid per callOffline, unlimited, microseconds per positionOffline, unlimited
Best whenYou want the data now, in a backtest or an agent, and want to cite itYou compute millions of positions or ship a desktop appYou want full control and have the time

Precision is not the difference

Swiss Ephemeris, Skyfield on a JPL kernel and this API all reproduce NASA/JPL to a small fraction of an arcsecond: far below anything a chart or a backtest can see. Choosing between them is about work, licence and provenance, not about decimals. Our own measurements are public: benchmark against JPL Horizons (12,400 comparisons, 1800-2200).

What you would have to build yourself

When to choose a library instead

If you need millions of positions a day, must run offline, or are shipping a desktop application, per-call pricing is the wrong tool: use Swiss Ephemeris (mind its AGPL or professional licence) or Skyfield. If you want a ready event calendar and Gann lines in a notebook today, with a hash you can cite, the API is the short path.

Data for your own analysis, never a trading signal or forecast.

Financial astrology data API · Examples · Live catalog and prices