LockLLM

What is LockLLM?

LockLLM is a complete AI security platform for LLM applications. It detects prompt injection, hidden instructions, and unsafe inputs before requests reach your model. It also provides custom content policies for content moderation, smart routing for cost optimization, and abuse protection.

What attacks does LockLLM detect?

LockLLM detects prompt injection (direct and indirect), jailbreak attempts, instruction and policy overrides, tool abuse, RAG and document injection, obfuscated or encoded attack patterns, and personally identifiable information (PII) in prompts.

Is LockLLM model-agnostic?

Yes. LockLLM is designed as a gateway that can sit in front of any LLM provider or self-hosted model. You send prompts to one API and get a risk signal back.

What is LockX?

LockX is LockLLM’s adaptive learning and policy loop used to improve detection quality over time while keeping integration simple.

Do you have benchmarks or evaluations?

Yes. LockLLM publishes benchmark and evaluation reporting on the homepage. Our evaluations show strong results and competitive performance versus comparable prompt-guard models on similar attack suites and methodology.

Do you have a free tier?

Yes. Safe scans are always free. Active users earn free monthly credits based on usage. Only pay small fees when threats are detected ($0.0001-$0.0002) or when smart routing saves costs (5% of savings). BYOK users can control costs with their own API keys across all major providers.

How do I use LockLLM?

You can scan prompts manually in the dashboard, or protect live traffic by routing requests through the LockLLM API, SDK, or Proxy using an API key.

Do I need to retrain my model?

No. LockLLM works independently of your AI model or AI agents and does not require retraining, fine-tuning, or prompt changes.