Corbits and NEAR AI are expanding their tie-up with IronClaw and provenance tooling to let teams verify where agent inference ran, what actions were taken and which controls applied.
Corbits and NEAR AI are pairing private inference, secure agent runtime controls and workflow provenance records to build more private and verifiable AI agent workflows for enterprise teams. The collaboration combines NEAR AI’s private inference technology with IronClaw, its open-source secure agent runtime, and Corbits’ shared workflow tooling so users can verify where AI inference took place, what an agent did and which security controls governed its actions. NEAR AI said supported models run inside Trusted Execution Environments, hardware-isolated areas designed to protect code and data during processing, with attestation reports and cryptographic signatures intended to provide evidence of how inference was handled. IronClaw adds WebAssembly sandboxes, capability-based permissions, endpoint allowlists and protected credential injection to limit what agents and connected tools can access. The update builds on Corbits’ previously announced July 15 integration of NEAR AI private inference infrastructure for enterprise agent workflows.