Common questions
How does detection work? Is it an AI model? +
Detection is deterministic rule-based pattern matching — not an LLM. That means <5ms latency, zero risk of the detector being itself injected, and no probabilistic false positives. Novel attack variants are added to the signature library on a rolling basis.
What happens when I hit my monthly limit? +
The API returns a 429 with a clear error. Your LLM calls won't break — you'll just need to handle the Injecto response before proceeding. Upgrade anytime from your dashboard.
Can Injecto catch novel or obfuscated attacks? +
The current version uses signature matching, which catches known patterns reliably. Obfuscated variants that don't match existing signatures may pass through. We update signatures regularly and an ML-enhanced tier is on the roadmap.
Do you store the prompts I send? +
Prompts are analyzed in memory and not persisted to disk. Only usage metadata (request count, timestamp) is stored against your API key.