One HTTP call returns structured recovery steps — backoff timings, retry counts, root cause. Built for autonomous agents that can't afford to stall.
# Your agent hits an OpenAI 429
curl -X POST https://resolve.arflow.io/resolve \
-H "Content-Type: application/json" \
-d '{
"service": "openai",
"error_code": "429",
"message": "Rate limit exceeded",
"agent_id": "my-agent-v2"
}'
↓ Response in ~50ms
{
"category": "rate_limit",
"confidence": 0.97,
"resolution": {
"backoff_seconds": [20, 40, 80, 160, 320],
"steps": ["Wait 20s before retry", ...]
}
}OpenAI 429, Postgres 53300, Stripe card_declined — your agent catches the error and sends it to Resolve in one HTTP call.
Three-tier matching: exact error code → regex pattern → best-for-service fallback. Returns in under 50ms.
Structured steps, backoff timings, retry counts, and documentation links — everything needed to recover and continue.
20 services. Growing weekly.
52 curated resolutions. 20 services. Live in under 5 minutes.