SWEN.AI Public API

Base URL

https://swen.ia.br/api/v1

Authentication

All endpoints work without a key at a low anonymous rate limit (10 requests/minute per IP), so you can try the API before requesting a key. For a higher limit (30 requests/minute), get a free key instantly below — no approval wait. Send it as a Bearer token:

Authorization: Bearer sk_swen_...

Keys are scoped per resource (news, models, tools, tutorials).

Endpoints

Example

curl https://swen.ia.br/api/v1/rankings/intelligence?limit=10 \
  -H "Authorization: Bearer sk_swen_..."

{
  "data": [
    { "rank": 1, "slug": "anthropic-claude-fable-5", "name": "Claude Fable 5", "vendor": "Anthropic", "score": 59.9, "url": "https://swen.ia.br/benchmark/anthropic-claude-fable-5" },
    { "rank": 2, "slug": "openai-gpt-5-6-sol-max", "name": "GPT-5.6 Sol (max)", "vendor": "OpenAI", "score": 58.9, "url": "https://swen.ia.br/benchmark/openai-gpt-5-6-sol-max" }
  ],
  "meta": { "count": 2, "source": "Artificial Analysis Intelligence Index", "updatedAt": "2026-07-17T14:30:00.000Z" }
}

Webhooks (push instead of polling)

Instead of polling /api/v1/news or /api/v1/models, register a webhook and get an HTTP POST the moment something changes. Requires an API key with the matching scope.

curl -X POST https://swen.ia.br/api/v1/webhooks \
  -H "Authorization: Bearer sk_swen_..." \
  -H "Content-Type: application/json" \
  -d '{"targetUrl": "https://your-site.com/webhooks/swen", "events": ["news.published"]}'

Response includes a secret — shown only once, used to verify every delivery. Available events: news.published, model.new, model.updated.

Each delivery includes X-SWEN-Timestamp (unix seconds) and X-SWEN-Signature (sha256=<hex>). To verify: compute HMAC-SHA256 of `${timestamp}.${rawBody}` using your secret, compare to the signature, and reject if the timestamp is more than ~5 minutes old (prevents replay of an intercepted delivery).

{
  "event": "news.published",
  "data": {
    "slug": "openai-lanca-gpt-5-6",
    "title": "OpenAI lança GPT-5.6",
    "summary": "...",
    "category": "Modelos",
    "publishedAt": "2026-07-22T22:00:00.000Z",
    "updatedAt": "2026-07-22T22:00:00.000Z",
    "source": "OpenAI",
    "url": "https://swen.ia.br/noticia/openai-lanca-gpt-5-6",
    "imageUrl": "https://swen.ia.br/...",
    "isManchete": true,
    "isBreakingNews": false,
    "isCover": true,
    "relevanceScore": 87,
    "tone": 4,
    "density": 3,
    "language": "pt-BR",
    "audioUrl": "https://swen.ia.br/audio/openai-lanca-gpt-5-6.mp3",
    "entities": [
      { "name": "OpenAI", "type": "company" },
      { "name": "GPT-5.6", "type": "product" }
    ]
  }
}

model.updated also includes a changes array with only the fields that actually changed (price, version, status, context window, featured):

{
  "event": "model.updated",
  "data": {
    "slug": "openai-gpt-5-6",
    "name": "GPT-5.6",
    "vendor": "OpenAI",
    "inputPricePerMillionTokens": 5,
    "tokensPerSecond": 120,
    "timeToFirstTokenMs": 300,
    "changes": [
      { "field": "inputPricePerMillionTokens", "before": 3, "after": 5 }
    ]
  }
}

If your endpoint doesn't respond with a 2xx, we retry automatically at 1 min, 5 min, and 30 min (X-SWEN-Retry header tells you the attempt number). If it still fails after 20 consecutive failed deliveries, the webhook is auto-disabled and we email the key owner — reactivate it with PATCH /api/v1/webhooks/{id} ({"active": true}).

Manage your webhooks with GET /api/v1/webhooks (list, no secret returned), PATCH /api/v1/webhooks/{id} (update targetUrl/events/filters/active without rotating the secret) and DELETE /api/v1/webhooks/{id}. Up to 10 active webhooks per key.

Test your endpoint anytime with POST /api/v1/webhooks/{id}/test — sends a sample payload (marked "test": true) right away, without waiting for a real event and without counting toward the auto-disable threshold.

Webhook filters — only breaking news, headlines, cover, or above a relevance score

By default a news.published webhook fires for every published article. Pass an optional filters object on creation to only get delivered the ones you care about — filtering happens on our side, so your endpoint never even receives the rest. Filters only apply to news.published (they have no effect on model.* events).

curl -X POST https://swen.ia.br/api/v1/webhooks \
  -H "Authorization: Bearer sk_swen_..." \
  -H "Content-Type: application/json" \
  -d '{
    "targetUrl": "https://your-site.com/webhooks/swen",
    "events": ["news.published"],
    "filters": { "onlyBreakingNews": true, "minRelevance": 80 }
  }'

All provided filters must match (AND) for a delivery to happen. Omit filters entirely to get everything.

Rate limits

• Anonymous (no key): 10 requests/minute per IP.
• Self-serve key (instant, below): 30 requests/minute.
• Negotiated partnership: higher limits on request.

Exceeding the limit returns 429 with a Retry-After header.

MCP Server (for AI agents)

SWEN.AI runs a public Model Context Protocol server exposing the same 4 data sources as callable tools — for Claude, ChatGPT, Cursor and any MCP-compatible agent. No API key required (IP rate limited).

https://swen.ia.br/api/mcp

Available tools: search_news, get_news_article, list_ai_models, get_model, get_intelligence_ranking, search_tools, list_tutorials.

Attribution

Intelligence ranking data is sourced from Artificial Analysis. If you display this data publicly, please credit SWEN.AI and link back to swen.ia.br.

Contract stability

All field names are English and versioned under /v1. We will never silently rename or remove a field within v1 — breaking changes ship as /v2.