# StableDetect API Reference

## Authentication

All endpoints require micropayment. Include payment headers as per the HTTP 402 specification. Payments are processed on Base mainnet (eip155:8453), Solana, or Tempo.

## Base URL

https://stabledetect.dev

All endpoints are relative to this base URL.

## Agent Workflow (Progressive)

1. Discover candidate endpoints with `mcp__agentcash__discover_api_endpoints("https://stabledetect.dev")`.
2. For selected POST endpoints, call `mcp__agentcash__check_endpoint_schema` before first fetch.
3. Execute with `mcp__agentcash__fetch`.

---

# AI Content Detection

Detect whether text, images, or video are AI-generated. Text detection is powered by [Pangram](https://www.pangram.com); image and video deepfake detection by [BitMind](https://bitmind.ai).

## POST /api/detect/text

Classify a text passage as AI-generated, AI-assisted, human-written, or mixed. Returns document-level fractions plus per-window segment labels with character offsets, so you can pinpoint exactly which parts are AI.

Price: $0.10 per started 1,000 words (a 3,500-word document costs $0.40). Call `mcp__agentcash__check_endpoint_schema` with your intended body for an exact quote.

Body:
- `text` (string, required) — the passage to analyze. Minimum 50 words, maximum 50,000 words. Under 50 words returns HTTP 400 before any charge.
- `publicDashboardLink` (boolean, optional, default false) — include a shareable pangram.com dashboard link for the analysis in the response.

Response (key fields):
```json
{
  "headline": "AI Detected",
  "prediction_short": "AI",
  "fraction_ai": 0.7,
  "fraction_ai_assisted": 0.2,
  "fraction_human": 0.1,
  "windows": [
    { "label": "AI-Generated", "ai_assistance_score": 0.85, "confidence": "High", "start_index": 0, "end_index": 812 }
  ]
}
```

`prediction_short` is one of `AI`, `AI-Assisted`, `Human`, `Mixed`. Fractions sum to ~1.0. Each window covers a slice of the input (`start_index`/`end_index` are character offsets).

## POST /api/detect/image

Detect whether an image is AI-generated or a deepfake.

Price: $0.02 per request

Body:
- `image` (string, required) — public http(s) URL or a data URL (`data:image/png;base64,...`). JPEG, PNG, GIF, BMP, WebP, AVIF.

Response:
```json
{ "isAI": true, "confidence": 0.87 }
```

`confidence` is 0-1; treat >= 0.7 as a strong signal either way.

## POST /api/detect/video

Detect whether a video is AI-generated or a deepfake. Accepts direct video URLs plus YouTube, Twitter/X, and TikTok links.

Price: $0.10 per request

Body:
- `video` (string, required) — video URL.
- `startTime` / `endTime` (numbers, optional, seconds) — analyze only a segment. Recommended for videos over ~2 minutes.
- `fps` (integer, optional) — frames per second to sample.

Response:
```json
{ "isAI": true, "confidence": 0.87 }
```

### Composition tips

- Verifying a social-media post: run /api/detect/text on the caption and /api/detect/image (or /video) on the media in parallel.
- Long articles: send the full text in one call — pricing scales per 1,000 words and windows tell you which sections are AI.
- Low-confidence results (0.4-0.6) are genuinely uncertain; report them as inconclusive rather than forcing a verdict.

---

# Pricing Summary

| Endpoint | Price |
|----------|-------|
| Text detection | $0.10 per started 1,000 words |
| Image detection | $0.02 |
| Video detection | $0.10 |