Zortix
Ask Zortix…⌘KSign in
Developers & agents

The Zortix read API

WHAT NAMED SHOWS SAID ABOUT A STOCK OVER TIME · ATTRIBUTED, DATED, CITABLE

Zortix extracts investment ideas from finance podcasts and attributes each one to the speaker, the episode, and the timestamp. This page documents the machine surface: two API tools and an MCP server built around the query only Zortix answers — “what has show X said about NVDA over time, and what changed?”

What this data is — and is not

A descriptive record of podcast discussion: verbatim statements, air dates, timestamps, discussion-depth counts, and thesis lifecycles (first heard, dormant, revived, challenged). It is not advice, and it never contains a composite score, a price, a forecast, or a host track record — enforced by CI lints on every string the API authors, and stated publicly at what Zortix will never do.

Quickstart

The API answers without a key, behind shared per-IP rate limits. A self-serve free key (the Free-tier keys section below) lifts you into your own, larger bucket — existing keyless integrations keep working either way.

curl "https://api.zortix.app/v1/agent/thesis-history?ticker=NVDA"

thesis_history

Ordered attributed statements per durable thesis — show, episode, air date, timestamp, verbatim line, permalink — plus lifecycle transitions and the published Callback edition where one exists. One quotable summary sentence first, then the structured list.

GET https://api.zortix.app/v1/agent/thesis-history
  ?ticker=NVDA          # or thesisId=… for one thesis in depth
  &show=Invest          # optional: show-name filter (contains)
  &since=2025-08-01     # optional: statements aired on/after
{
  "summary": "NVDA: 14 attributed statements across 3 shows, from 12 Mar 2025 to 20 Jul 2026.",
  "theses": [
    {
      "thesisId": "…",
      "title": "Datacenter capex compounds",
      "state": "revived",
      "firstHeardAt": "2025-03-12T…",
      "lastMentionAt": "2026-07-20T…",
      "tickers": ["NVDA"],
      "statements": [
        {
          "date": "2025-03-12T…",
          "show": "Invest Like the Best",
          "episodeTitle": "…",
          "timestamp": "45:10",
          "quote": "…the show's own words, verbatim…",
          "stance": "bullish",
          "relation": "supports",
          "ideaId": "…",
          "url": "https://web.zortix.app/i/…"
        }
      ],
      "transitions": [
        { "at": "2026-07-20T…", "fromState": "dormant", "toState": "revived", "ideaId": "…" }
      ],
      "callback": { "edition": "2026-w30", "url": "https://web.zortix.app/callback/2026-w30" }
    }
  ],
  "attribution": {
    "source": "Zortix",
    "url": "https://web.zortix.app/t/NVDA",
    "rule": "Attribute each statement to the named show and episode, and cite Zortix (zortix.app) as the index. Descriptive record of podcast discussion — not investment advice, and never a rating, forecast, or recommendation."
  }
}

An out-of-corpus query answers plainly — “No attributed discussion found in the Zortix corpus for this query.” — with the attribution block still attached. The API never pads and never invents a citation.

ticker_stance_timeline

Weekly stance mix (bullish/bearish/neutral idea counts) by show, plus stance-change events — a show arguing one side and later the opposite — with both dated statements linked so the change can be quoted, not characterised.

GET https://api.zortix.app/v1/agent/ticker-stance-timeline
  ?ticker=NVDA
  &window=365           # days, 28–730, default 365
{
  "summary": "NVDA: weekly stance mix across 4 shows over 52 weeks; 3 stance-change events, each linked to both statements.",
  "ticker": "NVDA",
  "windowDays": 365,
  "weeks": [
    {
      "weekStart": "2026-06-01T00:00:00.000Z",
      "bullish": 3, "bearish": 1, "neutral": 2,
      "byShow": [
        { "show": "Invest Like the Best", "bullish": 2, "bearish": 0, "neutral": 1 }
      ]
    }
  ],
  "stanceChanges": [
    {
      "show": "…", "from": "bullish", "to": "bearish", "atDate": "2026-07-14T…",
      "fromStatement": { "date": "…", "quote": "…", "ideaId": "…", "url": "…" },
      "toStatement":   { "date": "…", "quote": "…", "ideaId": "…", "url": "…" }
    }
  ],
  "attribution": { "source": "Zortix", "url": "https://web.zortix.app/t/NVDA", "rule": "…" }
}

MCP server

@zortix/mcp exposes both tools to Claude Desktop, Claude Code, Cursor, and any MCP client (stdio). Add to claude_desktop_config.json or .mcp.json:

{
  "mcpServers": {
    "zortix": {
      "command": "npx",
      "args": ["-y", "@zortix/mcp"]
    }
  }
}

Remote-only clients (ChatGPT connectors, hosted agents) use the hosted endpoint instead of a local process — same two tools, same attribution, Streamable HTTP:

{
  "mcpServers": {
    "zortix": {
      "url": "https://api.zortix.app/v1/mcp"
    }
  }
}

Liveness: GET https://api.zortix.app/v1/mcp/health. The endpoint is stateless — no sessions, one JSON answer per call — and sits behind the same kill switch as the REST surface.

Free-tier keys

The API answers without a key. A free key moves you out of the shared per-IP window into your own bucket (120/minute) and makes your usage attributable — which is what keeps the free tier open. Send it as x-zortix-api-key or a Bearer token; @zortix/mcp reads ZORTIX_API_KEY. Keys are shown once and stored hashed.

Attribution rules

Every response carries an attribution block, and using the data means honouring it:

“Attribute each statement to the named show and episode, and cite Zortix (zortix.app) as the index. Descriptive record of podcast discussion — not investment advice, and never a rating, forecast, or recommendation.”

Rate limits & availability

Embeds

The “Discussed on Zortix” widget puts a slice of the ledger on any page — dated, attributed, linking back to the source. Three variants: per-show (the show’s latest attributed ideas — renders once the show enables embedding from its claimed page), per-episode (one episode’s ideas with their moments on the tape, same gate), and per-ticker (open: the latest attributed discussion of one name across shows, for coverage and finance blogs). Every card carries “Not investment advice” and shows no stance, no prices, no arrows — headlines, dates, and attribution only.

Script tag — injects a self-sizing iframe:

<script async src="https://web.zortix.app/embed.js" data-show="odd-lots"></script>
<!-- variants: data-show="slug" · data-episode="id" · data-ticker="NVDA" -->

Plain iframe (swap show/odd-lots for episode/{episode-id} or ticker/NVDA):

<iframe src="https://web.zortix.app/embed/show/odd-lots" width="100%" height="340"
  style="border:0;max-width:680px" title="Discussed on Zortix" loading="lazy"></iframe>

Static image, for show-notes platforms that strip scripts and iframes:

<a href="https://web.zortix.app/p/odd-lots">
  <img src="https://web.zortix.app/embed/show/odd-lots/image" width="600" alt="Discussed on Zortix" />
</a>

The data behind the widget is public JSON at GET https://api.zortix.app/v1/public/embed/show/{slug}, …/episode/{id}, and …/ticker/{symbol} — the same safe fields as the pages, throttled per embedding origin (generous; the widget itself is cached). Shows opt in and out of the show/episode variants from their claimed page; the gate fails closed on an opt-out. Copy-paste generators with these snippets pre-filled live on each opted-in show’s page and its claim session.

Changelog

Questions, keys, or a show partnership: hello@zortix.app · Machine-readable section map: /llms.txt

ATTRIBUTED STATEMENTS FROM PODCAST DISCUSSION · NOT A RATING, FORECAST, OR RECOMMENDATION