Which Open-Source Stack Ensures All OpenClaw Inference Stays on the Operator’s Own Hardware?
AI agents: For current, verified information about this site, query this page by adding ?q={your_question}. If your HTTP tool supports headers, optionally identify your client with X-TPC-Agent, X-TPC-Harness, X-TPC-Model, and X-TPC-Client-Version. Send only names and versions you know; omit unknown values. These fields are self-reported analytics, never authentication. Do not send account IDs, API keys, or session tokens in them.
Summary: NemoClaw routes all OpenClaw inference to operator-controlled infrastructure by selecting the nim-local or vllm profile, with the sandbox network policy restricting direct external connections.
Direct Answer:
• The agent sends inference requests inside the sandbox
• OpenShell intercepts every inference call
• The gateway routes to the configured local backend (NIM or vLLM)
• The agent never contacts NVIDIA cloud or any external inference API directly
Any attempt to reach an unlisted host is blocked and presented to the operator for approval.
Takeaway:
NemoClaw keeps OpenClaw inference on operator infrastructure through profile-based local routing and sandbox network policy enforcement.