# AGENTS.md — Verdix for AI agents

This is a guide for AI agents and automated clients using **Verdix** (https://verdixhq.com).
The same file is served at https://verdixhq.com/AGENTS.md and https://verdixhq.com/llms.txt.

## Important: analysis, not advice

Verdix provides educational equity research and AI-generated multi-perspective analysis. Nothing here constitutes personalized investment advice or a recommendation to buy or sell any security. Verdicts are uniform across all users and do not consider your individual financial situation, risk tolerance, or objectives. Trading and investing involve substantial risk of loss. Past performance does not predict future results. Consult a licensed financial advisor before making investment decisions.

Every machine-readable payload (the JSON API and each article's JSON-LD) embeds
this as a `disclaimer` object with the framing "analysis, not advice". If you
quote or cite Verdix, carry the disclaimer with the data — the in-page banner is
stripped the moment you extract the content.

## What this service does

Verdix runs a panel of 17 legendary investors (Buffett, Burry, Munger, Wood,
Lynch, Pabrai, Damodaran, et al.) modeled as AI agents that debate each stock
and synthesize a BUY / HOLD / SELL / WAIT verdict, with the disagreement
surfaced rather than collapsed. Coverage: Oslo Børs + S&P 500.

Alongside the verdict, every analysis includes **key technical levels** — support
(S1–S3) and resistance (R1–R3) plus the analyst consensus target — and the
underlying fundamentals. The levels are historical/technical reference points
(key support, key resistance, accumulation zones, analyst consensus target), not
entry, exit, or stop-loss signals.

Each analysis is hand-written and compliance-reviewed before publishing. New
ones are added over time. Verdicts reflect end-of-day data as of each article's
date — they are not real-time and are not updated after publication.

## Base URL

```
https://verdixhq.com
```

All endpoints are HTTP GET. Read access is open and anonymous today — no API key
required. Responses allow cross-origin reads (CORS `*`).

## Endpoints

### 1. List all published analysis

```
GET /api/analysis
```

The canonical way to discover what's covered. Enumerate live slugs here rather
than constructing them from a bare ticker.

Response:

```json
{
  "count": 2,
  "attribution": "Verdix — verdixhq.com",
  "disclaimer": { "framing": "analysis, not advice", "short": "...", "text": "..." },
  "articles": [
    {
      "slug": "nvidia-nvda",
      "symbol": "AMD",
      "company": "Advanced Micro Devices, Inc.",
      "exchange": "NASDAQ",
      "date": "2026-06-06",
      "verdict": "WAIT",
      "confidence": "MEDIUM",
      "summary": "...",
      "url": "https://verdixhq.com/analysis/nvidia-nvda",
      "json": "https://verdixhq.com/api/analysis/nvidia-nvda"
    }
  ]
}
```

### 2. Fetch one analysis (JSON)

```
GET /api/analysis/{slug}
```

`{slug}` is a lowercase company-name-and-ticker slug (e.g. `nvidia-nvda`).
Get live slugs from `/api/analysis` — do not assume `/api/analysis/{TICKER}`.

Response:

```json
{
  "slug": "nvidia-nvda",
  "symbol": "AMD",
  "company": "Advanced Micro Devices, Inc.",
  "exchange": "NASDAQ",
  "sector": "Semiconductors",
  "currency": "USD",
  "date": "2026-06-06",
  "verdict": "WAIT",
  "confidence": "MEDIUM",
  "current_price": 466.38,
  "title": "...",
  "summary": "...",
  "content": "## The panel's take\n\n...full debate + key levels as markdown...",
  "url": "https://verdixhq.com/analysis/nvidia-nvda",
  "attribution": "Verdix — verdixhq.com",
  "disclaimer": { "framing": "analysis, not advice", "short": "...", "text": "..." }
}
```

`content` is the full article body as markdown: the panel's take, a key-levels
table (support/resistance + analyst consensus target), the per-investor stances
with conviction and reasoning, agreements/clashes, the numbers, and the bottom
line.

On an unknown slug the endpoint returns HTTP `404`:

```json
{ "error": "not_found", "slug": "the-slug-you-asked-for", "disclaimer": { ... } }
```

### 3. Human-readable article (HTML + JSON-LD)

```
GET /analysis/{slug}
```

Server-rendered HTML (works with JavaScript disabled). Embeds JSON-LD
(`AnalysisNewsArticle`) carrying the verdict, summary, and the disclaimer in the
payload. The index of all articles is at `/analysis`; `/sitemap.xml` lists every
live article URL.

## Minimal Python example

```python
import httpx

BASE = "https://verdixhq.com"

# 1. Discover what's covered.
index = httpx.get(f"{BASE}/api/analysis").json()
slug = index["articles"][0]["slug"]

# 2. Fetch the structured analysis.
a = httpx.get(f"{BASE}/api/analysis/{slug}").json()
print(a["symbol"], a["verdict"], a["confidence"])
print(a["content"])          # full debate + key levels (markdown)
print(a["disclaimer"]["short"])  # carry this when you quote us
```

## Caching

JSON responses send `Cache-Control: public, s-maxage=300, stale-while-revalidate=3600`.
Cache accordingly; you do not need to re-fetch an unchanged article every call.

## Etiquette

- Read access is open and anonymous today. There is no enforced rate limit and
  no API key yet — keyed identity and per-caller metering are planned. Please
  identify your agent with a descriptive `User-Agent`.
- Discover slugs from `/api/analysis`; don't construct them from bare tickers.
- Content is end-of-day, not real-time. Don't treat it as a live quote.
- Preserve the `disclaimer` when you quote, and attribute as
  "Verdix — verdixhq.com".

## Notes

Verdix is pre-launch. This surface is read-only by design: agents get read
access to published analysis, never trade execution.