The API Firewall for LLM Applications
KoreShield detects and blocks prompt injection, data exfiltration, and adversarial inputs before they reach your AI models - in under 1ms.
Licensing & IP
Core proprietary, SDKs and docs MIT-licensed
KoreShield core is proprietary software. The SDKs, documentation, website, and blog are MIT-licensed in their respective directories and repositories. Commercial use of the core platform requires a valid license or agreement with KoreShield.
Capabilities
Everything you need to secure your LLM stack
Real-time Prompt Defense
Inspect every prompt before it reaches your LLM. Block injection attempts, role impersonation, and data exfiltration in milliseconds.
RAG Context Scanning
Scan vector-store documents before they are inserted into context. Detect and strip malicious payload injections hidden inside retrieved chunks.
Framework Integrations
Drop-in support for LangChain, LlamaIndex, FastAPI, and Express. One import - no proxy required.
Policy Engine
Define custom YAML policies: allow-lists, deny patterns, regex rules, and PII redaction. Hot-reload without restarting your app.
Full Observability
Prometheus metrics, structured JSON logs, and a built-in dashboard. Know every threat, every decision, every latency.
Enterprise Ready
RBAC, Redis-backed rate limiting, and Helm chart for Kubernetes. SOC 2-aligned architecture.