NEAR AI Launches IronClaw, Confidential GPU Marketplace, and Multimodal Confidential Inference

Today at NEARCON 2026, NEAR AI introduced two major products—IronClaw and the Confidential GPU Marketplace—along with significant new platform capabilities including multimodal support and flexible privacy controls.

Together, these launches establish a new standard for agent-native infrastructure: hardware-enforced security, runtime verification, and confidentiality by design.

As AI systems evolve from chat interfaces to always-on agents that hold credentials, call APIs, and execute transactions autonomously, infrastructure must move from policy-based trust to cryptographic guarantees. NEAR AI is building that foundation.

IronClaw: Always-On Agents, Hardware-Enforced Security

IronClaw is a secure, open-source AI agent runtime built in Rust and deployed inside encrypted Trusted Execution Environments (TEEs) on NEAR AI Cloud.

It carries forward the original OpenClaw vision—autonomous agents that can act, transact, and coordinate on your behalf—while strengthening it with deeper, cryptographic security. The goal remains the same: make always-on agents practical. The difference is that security is enforced at the architectural layer.

IronClaw isolates credentials in an encrypted vault, sandboxes tools in WebAssembly containers, and ensures secrets never reach the model. Prompt injection and credential exfiltration are mitigated by design, not by policy. Jobs run inside hardware-backed enclaves with verifiable attestation—security that does not depend on operator trust.

For AI developers and teams exploring OpenClaw-style agents, IronClaw is designed to feel familiar—without inheriting the security and privacy tradeoffs. Migration is straightforward, local development is supported, and production deployments come with enterprise-grade reliability and SLAs for teams that need them.

The result: always-on agents that can browse, transact, and automate—with confidentiality enforced at the hardware layer.

→ ironclaw.com

Confidential GPU Marketplace: Confidential Inference Without Building Confidential Computing Infrastructure

NEAR AI’s Confidential GPU Marketplace is the first TEE-secured compute network designed for enterprise and government AI workloads, built on NEAR DCML.

Jobs execute inside hardware-enforced encrypted enclaves where data is protected in memory. The host OS, the GPU operator, and even NEAR AI are cryptographically locked out. Every request returns hardware-signed attestation in under 30 seconds.

The Confidential GPU Marketplace is designed to solve problems for several sides of the current enterprise AI market. Enterprises need secure AI inference. Public clouds, however, allow operators to access data in transit, disqualifying them for PHI, PII, and classified workloads. Self-hosting confidential infrastructure costs $5M+ and six months before serving a single request. Meanwhile, GPU providers operate at 30–40% idle capacity—but without confidential computing, regulated workloads cannot be deployed.

The Confidential GPU Marketplace now connects both sides.

Providers monetize idle GPUs with enterprise workloads running inside encrypted TEEs. Customers receive cryptographic proof their data was never exposed—no infrastructure to build, no compliance liability to absorb.

→ near.ai/gpu-marketplace

Multimodal + Multiprivacy: One Platform, Multiple Modalities, Flexible Confidentiality

NEAR AI now supports multimodal inputs—text, image, and voice—within a unified execution environment.

Developers can run fully confidential models inside TEEs or access models anonymously through the NEAR AI gateway. Whether workloads require strict isolation or lightweight private access, confidentiality is enforced at the hardware layer.

This flexibility allows agents and applications to choose the appropriate privacy model without sacrificing performance or usability.

Unified Infrastructure for a User-Owned AI Economy

These launches reflect a unified direction for NEAR AI.

IronClaw secures the agent runtime, ensuring credentials and tools operate inside isolated, hardware-enforced environments. The Confidential GPU Marketplace secures model execution, enabling sensitive workloads to run inside encrypted enclaves with verifiable attestation. Multimodal and multiprivacy capabilities extend that same security model across text, image, and voice—with flexible confidentiality at the execution layer.

Together, these launches form a unified stack where agents can compute, transact, and coordinate without exposing secrets or relying on policy-based assurances. As AI systems become economically autonomous, infrastructure must internalize confidentiality, verification, and control at the protocol level. NEAR AI is pioneering this infrastructure to ensure the fast-growing AI economy is user-owned.

Explore the full stack at near.ai.

Hector Martinez

In today’s interconnected world, health challenges are global—and so are the solutions. The Global Health Connect Podcast explores the intersection of global healthcare and innovative partnerships. Join us as we uncover the stories.