NEAR AI announced the launch of IronClaw, an open-source and verifiable AI agent runtime that carries forward the original OpenClaw vision, while strengthening it with cryptographic security. Built in Rust and deployed inside an encrypted Trusted Execution Environment (TEE) on NEAR AI Cloud, IronClaw is a security-focused AI agent runtime designed to enable autonomous agents with built-in privacy safeguards and controlled access to user assets and data.
Designed with user privacy and security at its core, IronClaw delivers a privacy-first experience for users who want the power of modern AI agents while ensuring sensitive data and execution remain under the user’s direct control. IronClaw has launched with a free Starter tier that includes one AI agent instance hosted within NEAR’s secure environment and powered by NEAR AI’s inference infrastructure, with flexible paid tiers available for users who wish to deploy and scale additional agents.
Security is foundational to IronClaw’s design. Built with a defense-in-depth approach, IronClaw layers multiple protections to safeguard data and prevent misuse, allowing AI agents to execute tasks, access tools, and maintain memory over time within a controlled, security-hardened environment. All third-party or untrusted tools run inside isolated, sandboxed environments, ensuring they can only access the specific resources they are explicitly permitted to use. Network requests are also tightly controlled and limited to approved destinations, while sensitive credentials are securely injected at runtime and never exposed directly to tools.
Illia Polosukhin, Co-Founder of NEAR Protocol and Founder of NEAR AI, said, “For AI to become a real extension of the human experience, it must exist within an architecture that can be trusted, and that is what we built with IronClaw. It’s an agentic harness designed for security, similar to the chain signatures infrastructure we’ve already built on NEAR. It brings the same full-stack approach to the AI layer, combining private and confidential inference with built-in security, laying the foundation for an AI system that ensures privacy and security by default.”
IronClaw addresses persistent issues around data leaks and misuse that exist currently with AI agents, continuously scanning agent activity to prevent data exploitation. IronClaw enforces strict limits on resource usage and applies advanced protections against prompt injection attacks. All data remains stored locally within a user’s PostgreSQL database with personal data encrypted using industry standard AES-256-GCM encryption. To bolster transparency and control for the user, IronClaw maintains a comprehensive audit log of all tool activity, while collecting no telemetry or analytics data to ensure data never leaves the user’s local environment.
George Zeng, Chief Product Officer and General Manager of NEAR AI, said, “AI agents are rapidly becoming part of core workflows, and critical questions around security, compliance, and data control remain unresolved. We’re extending NEAR AI’s commitment to secure, privacy-first infrastructure with IronClaw, giving developers and enterprises the foundation they need to deploy agentic AI systems responsibly without sacrificing control or transparency.”
In a world where AI systems are increasingly aligned with corporate interests and opaque about data handling, IronClaw ensures that your data remains protected through local storage and control, transparency by design, and a layered security framework. Privacy and security are not add-on features, they are intrinsic to IronClaw’s design. IronClaw is available to deploy now, and code can be accessed on NEAR AI’s GitHub.
To learn more about NEAR AI, visit https://near.ai/