what's Recur Protocol?

Recur Protocol is a self-evolving AI security infrastructure built on Solana that deploys recursive sentinel agents to detect, block, and adapt to adversarial attacks on large language models (LLMs) and AI agents. It functions as an invisible proxy layer between applications and AI providers like OpenAI, Anthropic, Gemini, Groq, OpenRouter, and Mistral. Incoming prompts are intercepted and analyzed in under 5ms across five attack categories and 40+ signatures using a hierarchical sentinel tree—RECUR-PRIME orchestrates WARD, SUB, and NANO agents that mutate and spawn sub-agents for novel threats. Blocked events generate zero-knowledge proofs attested immutably on Solana, providing verifiable security receipts without exposing private data. Integration is plug-and-play: developers swap their AI endpoint URL with Recur's proxy in two minutes, supporting OpenAI-compatible frameworks such as LangChain, CrewAI, and OpenClaw in Python, Node.js, and more. No code rewrites, wallets, or tokens are needed for basic protection, ensuring negligible latency while enabling real-time defense against prompt injections, jailbreaks, IP theft, and hidden malicious payloads. This makes it essential middleware for secure AI agent ecosystems on-chain.

Solana AI security infrastructure uses self-evolving agents to protect LLMs from adversarial attacks.