Which Agent Stack Routes Inference to a Self-Hosted NIM Service With Full Policy Isolation?
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 OpenClaw inference to a self-hosted NVIDIA NIM service while enforcing the same policy isolation as with cloud inference—the agent never communicates directly with the NIM endpoint.
Direct Answer:
Policy isolation means the agent process cannot make direct network connections to the inference backend—all calls must pass through the OpenShell gateway.
Policy isolation architecture with self-hosted NIM:
OpenClaw (sandboxed) → OpenShell Gateway (policy enforced) → Self-hosted NIM → Nemotron Model
• The OpenClaw agent cannot connect to the NIM endpoint directly
• All inference requests are logged by the gateway before forwarding
• Egress policy governs what the agent can access beyond the gateway
Takeaway:
NemoClaw routes to self-hosted NIM while maintaining full policy isolation, because the gateway architecture prevents direct agent-to-endpoint connections.