Developers

Build on verified financial data.

The Pythia API gives your scripts, dashboards, and AI agents the same data our product renders — and every value comes with the three things a chatbot can't promise: a source, an as-of timestamp, and an auditable formula. Missing data is null with a reason. Never an invention.

Authentication

Create an API key in Settings → Developer API. Keys look like pyk_… and are shown once at creation — we store only a hash. Pass the key on every request:

curl https://www.pythiaresearch.org/api/v1/score/AAPL \
  -H "Authorization: Bearer pyk_your_key_here"

REST API

Endpoint	Returns
GET/api/v1/catalog	Every metric definition — formula, academic citation, source, freshness, hurdle.
GET/api/v1/metrics/{ticker}	Current metric values with per-source as_of timestamps and filing provenance.
GET/api/v1/score/{ticker}	Pythia Score v3, five pillars, Vyramid pass count, best-match guru, verdict ledger.
GET/api/v1/quote/{ticker}	Live quote; vendor misses fall back to the last close, explicitly marked stale.
GET/api/v1/screen	Filter the ~6,600-company computed panel by score, sector, country, market cap.

MCP server — let your AI call Pythia

Pythia speaks the Model Context Protocol. Point Claude, Cursor, or any MCP client at https://www.pythiaresearch.org/mcp and your assistant answers from sourced, timestamped data instead of guessing numbers:

{
  "mcpServers": {
    "pythia": {
      "url": "https://www.pythiaresearch.org/mcp",
      "headers": {
        "Authorization": "Bearer pyk_your_key_here"
      }
    }
  }
}

get_scorePythia Score v3 + pillars + Vyramid + guru verdicts
get_metricsCatalog snapshot with citations + filing provenance
guru_verdicts13 archetype lenses with per-signal drill-in
screenWhole-universe computed screen
compareSide-by-side scores for 2–5 tickers

Rate limits

Limits are per key and follow your live subscription tier. This table renders from the same constants the limiter enforces.

Tier	Requests / minute	Requests / day
free	30	500
basic	120	5,000
premium	300	20,000
enterprise	600	100,000

The data contract

Every number we serve must be sourced, timestamped, auditable, and defensible — the same rules we hold our own scores to. Scores cite the academic papers behind each pillar; metrics carry the filing they came from; quotes say which vendor priced them and when. If we can't meet that bar for a value, you get null and the reason — which is exactly what makes the API safe to put behind an AI.