Question 1

What is Secra?

Accepted Answer

Secra is an AI security infrastructure layer that sits between your AI agents and the LLM. It detects and blocks prompt injection attacks, persona hijacking, and data exfiltration in real time before they reach your model.

Question 2

How does Secra detect prompt injection?

Accepted Answer

Secra uses a 3-layer detection pipeline: Layer 0 is an Aho-Corasick pre-processor that catches 30+ known injection signatures in under 1ms at zero token cost. Layer 1 applies a regex rule engine for pattern-based threats. Layer 2 uses Groq LLM inference only for ambiguous cases scoring 0.25-0.75.

Question 3

Is Secra free to use?

Accepted Answer

Yes. Secra's Free plan includes 500K tokens per month at no cost, forever. Paid plans start at $15/month for 5M tokens (Developer) and $49/month for 50M tokens with real-time monitoring (Pro).

Question 4

Does Secra store my prompts?

Accepted Answer

No. Secra never stores raw prompts. Only SHA-256 hashes and metadata (scan type, threat score, recommendation, token count) are retained for audit history.

Question 5

How do I integrate Secra into my AI agent?

Accepted Answer

Install the SDK with 'pip install secra-sdk' (Python) or 'npm install @secra/sdk' (JavaScript), set your SECRA_API_KEY environment variable, and call client.scan(prompt) before passing input to your LLM. The entire integration takes under 2 minutes.