fix(policy): add openclaw-pricing preset and auto-suggest for OpenClaw

docs/network-policy/customize-network-policy.mdx

@@ -179,6 +179,7 @@ Available presets:
 | `jira` | Atlassian Jira API |
 | `local-inference` | Local Ollama and vLLM through the host gateway |
 | `npm` | npm and Yarn registries |
+| `openclaw-pricing` | OpenClaw model-pricing reference fetch (LiteLLM and OpenRouter) |
 | `outlook` | Microsoft 365 and Outlook |
 | `pypi` | Python Package Index |
 | `slack` | Slack API and webhooks |

docs/network-policy/integration-policy-examples.mdx

@@ -60,6 +60,7 @@ NemoClaw ships maintained policy presets for common services in `nemoclaw-bluepr
 | Hugging Face Hub and Inference API | `huggingface` |
 | Jira and Atlassian Cloud | `jira` |
 | Local Ollama or vLLM through the host gateway | `local-inference` |
+| OpenClaw model-pricing reference fetch | `openclaw-pricing` |
 | npm and Yarn packages | `npm` |
 | Microsoft 365, Outlook, and Graph API | `outlook` |
 | Python Package Index | `pypi` |
@@ -252,6 +253,23 @@ $ nemoclaw my-assistant exec -- brew install <formula>
 
 You do not need to bootstrap Homebrew, install build dependencies, or source `brew shellenv` inside the sandbox.
 
+## Model Pricing
+
+OpenClaw's gateway fetches reference pricing from LiteLLM and OpenRouter on every start so it can populate `usage.cost` in session JSONL records.
+The default-strict egress policy denies both hosts.
+The fetch fails closed, the gateway logs `[gateway/model-pricing] LiteLLM pricing fetch failed: TypeError: fetch failed` (and the matching OpenRouter line) on every startup, and every session record records `usage.cost = 0` even though the input and output token counts populate correctly.
+Tools that read the session log to display per-turn cost (audit dashboards, compliance review surfaces) cannot distinguish a real free run from this silent failure.
+
+Apply the `openclaw-pricing` preset to allow both pricing endpoints.
+The preset pins each host to a single read-only path so it does not widen egress beyond the pricing fetch:
+
+```console
+$ nemoclaw my-assistant policy-add openclaw-pricing --dry-run
+$ nemoclaw my-assistant policy-add openclaw-pricing --yes
+```
+
+After the next gateway restart the WARN entries stop and `usage.cost` populates from the fetched pricing tables.
+
 ## Local Inference
 
 Use `local-inference` when the sandbox needs access to host-side local inference services such as Ollama or vLLM through the OpenShell host gateway.

nemoclaw-blueprint/policies/presets/openclaw-pricing.yaml

@@ -0,0 +1,39 @@
+# SPDX-FileCopyrightText: Copyright (c) 2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# SPDX-License-Identifier: Apache-2.0
+# OpenClaw's gateway/model-pricing subsystem fetches reference pricing from
+# LiteLLM and OpenRouter on every gateway start. With default-strict egress
+# both fetches fail closed with `TypeError: fetch failed`, the gateway
+# emits a WARN every startup, and every session JSONL row records
+# `usage.cost = 0` even though token counts populate correctly. Tools
+# reading the session log to display per-turn cost (audit dashboards,
+# compliance review surfaces) cannot distinguish a real free run from
+# the silent failure mode. Enabling this preset allowlists the two
+# specific pricing endpoints, pinned to GET-only on a single path each.
+preset:
+  name: openclaw-pricing
+  description: "OpenClaw model-pricing reference fetch (LiteLLM + OpenRouter)"
+network_policies:
+  openclaw-pricing:
+    name: openclaw-pricing
+    endpoints:
+      # LiteLLM publishes the upstream pricing table on raw.githubusercontent.com.
+      # Path is pinned to the single JSON file the fetch reads, GET only.
+      - host: raw.githubusercontent.com
+        port: 443
+        protocol: rest
+        enforcement: enforce
+        rules:
+          - allow:
+              method: GET
+              path: "/BerriAI/litellm/main/model_prices_and_context_window.json"
+      # OpenRouter exposes its model catalogue (with cost fields) under
+      # /api/v1/models. The fetch is read-only, so POST is not allowed.
+      - host: openrouter.ai
+        port: 443
+        protocol: rest
+        enforcement: enforce
+        rules:
+          - allow: { method: GET, path: "/api/v1/models" }
+    binaries:
+      - { path: /usr/local/bin/node }
+      - { path: /usr/bin/node }

src/lib/onboard/machine/handlers/policies.test.ts

@@ -200,4 +200,26 @@ describe("handlePoliciesState", () => {
       expect.objectContaining({ selectedPresets: ["npm", "github"] }),
     );
   });
+
+  it("forwards 'openclaw' to setupPoliciesWithSelection when agent is null (default OpenClaw)", async () => {
+    const { deps, calls } = createDeps();
+
+    await handlePoliciesState({ ...baseOptions(deps), agent: null });
+
+    expect(calls.setupPolicies).toHaveBeenCalledWith(
+      "my-assistant",
+      expect.objectContaining({ agent: "openclaw" }),
+    );
+  });
+
+  it("forwards 'hermes' to setupPoliciesWithSelection when agent.name is hermes", async () => {
+    const { deps, calls } = createDeps();
+
+    await handlePoliciesState({ ...baseOptions(deps), agent: { name: "hermes" } });
+
+    expect(calls.setupPolicies).toHaveBeenCalledWith(
+      "my-assistant",
+      expect.objectContaining({ agent: "hermes" }),
+    );
+  });
 });

src/lib/onboard/machine/handlers/policies.ts

@@ -75,6 +75,7 @@ export interface PoliciesStateOptions<Agent, WebSearchConfig> {
         disabledChannels?: string[] | null;
         webSearchConfig: WebSearchConfig | null;
         provider: string;
+        agent?: string | null;
         webSearchSupported: boolean;
         hermesToolGateways: string[];
         onSelection: (policyPresets: string[]) => void;
@@ -177,6 +178,11 @@ export async function handlePoliciesState<Agent, WebSearchConfig>({
       disabledChannels: activeSandbox?.disabledChannels,
       webSearchConfig,
       provider,
+      // selectOnboardAgent returns null for the default OpenClaw path (no
+      // --agent flag, no recorded agent). Normalise that to "openclaw" so
+      // openclaw-pricing auto-suggestion still fires; explicit Hermes runs
+      // keep their own name and are left alone.
+      agent: (agent as { name?: string } | null)?.name ?? "openclaw",
       webSearchSupported,
       hermesToolGateways,
       onSelection: (policyPresets) => {

src/lib/onboard/policy-presets.ts

@@ -12,20 +12,25 @@ export interface SuggestedPolicyPresetOptions {
   enabledChannels?: string[] | null;
   webSearchConfig?: WebSearchConfig | null;
   provider?: string | null;
+  agent?: string | null;
   isNonInteractive?: () => boolean;
 }
 
 export function getSuggestedPolicyPresets({
   enabledChannels = null,
   webSearchConfig = null,
   provider = null,
+  agent = null,
   isNonInteractive,
 }: SuggestedPolicyPresetOptions = {}): string[] {
   const suggestions = ["pypi", "npm"];
 
   if (provider && LOCAL_INFERENCE_PROVIDERS.includes(provider)) {
     suggestions.push("local-inference");
   }
+  if (agent === "openclaw") {
+    suggestions.push("openclaw-pricing");
+  }
   const usesExplicitMessagingSelection = Array.isArray(enabledChannels);
   const nonInteractive =
     isNonInteractive?.() ?? process.env.NEMOCLAW_NON_INTERACTIVE === "1";

src/lib/onboard/policy-selection.ts

@@ -37,6 +37,7 @@ export type SetupPresetSuggestionOptions = {
   enabledChannels?: string[] | null;
   webSearchConfig?: WebSearchConfig | null;
   provider?: string | null;
+  agent?: string | null;
   knownPresetNames?: string[] | null;
   webSearchSupported?: boolean | null;
   hermesToolGateways?: string[] | null;
@@ -48,6 +49,7 @@ export type SetupPolicySelectionOptions = {
   webSearchConfig?: WebSearchConfig | null;
   enabledChannels?: string[] | null;
   provider?: string | null;
+  agent?: string | null;
   knownPresetNames?: string[];
   webSearchSupported?: boolean | null;
   hermesToolGateways?: string[] | null;
@@ -127,7 +129,12 @@ export function computeSetupPresetSuggestions(
   tierName: string,
   options: SetupPresetSuggestionOptions = {},
 ): string[] {
-  const { enabledChannels = null, webSearchConfig = null, provider = null } = options;
+  const {
+    enabledChannels = null,
+    webSearchConfig = null,
+    provider = null,
+    agent = null,
+  } = options;
   const known = Array.isArray(options.knownPresetNames) ? new Set(options.knownPresetNames) : null;
   const supportOptions = { webSearchSupported: options.webSearchSupported };
   const suggestions = deps.tiers
@@ -144,6 +151,7 @@ export function computeSetupPresetSuggestions(
   };
   if (webSearchConfig) add("brave");
   if (provider && deps.localInferenceProviders.includes(provider)) add("local-inference");
+  if (agent === "openclaw") add("openclaw-pricing");
   if (Array.isArray(enabledChannels)) {
     for (const channel of enabledChannels) add(channel);
     for (const preset of requiredMessagingChannelPolicyPresets(enabledChannels)) add(preset);
@@ -237,6 +245,7 @@ export async function setupPoliciesWithSelection(
   const webSearchConfig = options.webSearchConfig || null;
   const enabledChannels = Array.isArray(options.enabledChannels) ? options.enabledChannels : null;
   const provider = options.provider || null;
+  const agent = options.agent || null;
   const hermesToolGateways = Array.isArray(options.hermesToolGateways)
     ? options.hermesToolGateways
     : null;
@@ -315,6 +324,7 @@ export async function setupPoliciesWithSelection(
       enabledChannels,
       webSearchConfig,
       provider,
+      agent,
       knownPresetNames: allPresets.map((preset) => preset.name),
       webSearchSupported: options.webSearchSupported,
       hermesToolGateways,

test/onboard-policy-suggestions.test.ts

@@ -14,6 +14,7 @@ const {
       enabledChannels?: string[] | null;
       knownPresetNames: string[];
       provider?: string | null;
+      agent?: string | null;
       webSearchConfig?: { fetchEnabled?: boolean; provider?: string | null } | null;
       webSearchSupported?: boolean | null;
     },
@@ -25,6 +26,7 @@ const {
   getSuggestedPolicyPresets: (options?: {
     enabledChannels?: string[] | null;
     provider?: string | null;
+    agent?: string | null;
   }) => string[];
 };
 
@@ -103,6 +105,30 @@ describe("onboard policy preset suggestions", () => {
     expect(getSuggestedPolicyPresets({})).not.toContain("local-inference");
   });
 
+  it("suggests openclaw-pricing preset only for the openclaw agent", () => {
+    expect(getSuggestedPolicyPresets({ agent: "openclaw" })).toContain("openclaw-pricing");
+    expect(getSuggestedPolicyPresets({ agent: "hermes" })).not.toContain("openclaw-pricing");
+    expect(getSuggestedPolicyPresets({ agent: null })).not.toContain("openclaw-pricing");
+    expect(getSuggestedPolicyPresets({})).not.toContain("openclaw-pricing");
+  });
+
+  it("adds openclaw-pricing to tier suggestions when agent is openclaw", () => {
+    const knownWithPricing = [...known, "openclaw-pricing"];
+    const openclawSuggestions = computeSetupPresetSuggestions("balanced", {
+      enabledChannels: [],
+      knownPresetNames: knownWithPricing,
+      agent: "openclaw",
+    });
+    expect(openclawSuggestions).toContain("openclaw-pricing");
+
+    const hermesSuggestions = computeSetupPresetSuggestions("balanced", {
+      enabledChannels: [],
+      knownPresetNames: knownWithPricing,
+      agent: "hermes",
+    });
+    expect(hermesSuggestions).not.toContain("openclaw-pricing");
+  });
+
   it("returns balanced tier defaults without messaging presets when no channels enabled", () => {
     const suggestions = computeSetupPresetSuggestions("balanced", {
       enabledChannels: [],

test/policies.test.ts

@@ -142,9 +142,9 @@ selectFromList(items, options)
 
 describe("policies", () => {
   describe("listPresets", () => {
-    it("returns all 19 presets", () => {
+    it("returns all 20 presets", () => {
       const presets = policies.listPresets();
-      expect(presets.length).toBe(19);
+      expect(presets.length).toBe(20);
     });
 
     it("each preset has name and description", () => {
@@ -173,6 +173,7 @@ describe("policies", () => {
         "nous-image",
         "nous-web",
         "npm",
+        "openclaw-pricing",
         "outlook",
         "pypi",
         "slack",
@@ -313,6 +314,45 @@ describe("policies", () => {
       }
     });
 
+    it("openclaw-pricing preset pins LiteLLM and OpenRouter reference fetches to GET-only paths", () => {
+      // OpenClaw's gateway/model-pricing subsystem fetches the LiteLLM
+      // pricing table and the OpenRouter model catalogue on every start.
+      // Both endpoints are read-only metadata fetches, so the preset must
+      // expose exactly one GET rule per host on the specific path each
+      // fetch reads, with no wildcards that could widen into a general
+      // raw.githubusercontent.com or openrouter.ai escape hatch.
+      const parsed = parsePresetYaml("openclaw-pricing");
+      const endpoints: Array<Record<string, unknown>> =
+        parsed?.network_policies?.["openclaw-pricing"]?.endpoints ?? [];
+      expect(endpoints).toHaveLength(2);
+
+      const litellm = endpoints.find((item) => item.host === "raw.githubusercontent.com");
+      if (!litellm) throw new Error("expected raw.githubusercontent.com endpoint");
+      expect(litellm.port).toBe(443);
+      expect(litellm.protocol).toBe("rest");
+      expect(litellm.enforcement).toBe("enforce");
+      expect(litellm.rules).toEqual([
+        {
+          allow: {
+            method: "GET",
+            path: "/BerriAI/litellm/main/model_prices_and_context_window.json",
+          },
+        },
+      ]);
+
+      const openrouter = endpoints.find((item) => item.host === "openrouter.ai");
+      if (!openrouter) throw new Error("expected openrouter.ai endpoint");
+      expect(openrouter.port).toBe(443);
+      expect(openrouter.protocol).toBe("rest");
+      expect(openrouter.enforcement).toBe("enforce");
+      expect(openrouter.rules).toEqual([{ allow: { method: "GET", path: "/api/v1/models" } }]);
+
+      const binaries: Array<{ path: string }> =
+        parsed?.network_policies?.["openclaw-pricing"]?.binaries ?? [];
+      const binaryPaths = binaries.map((entry) => entry.path).sort();
+      expect(binaryPaths).toEqual(["/usr/bin/node", "/usr/local/bin/node"]);
+    });
+
     it("local-inference preset includes openclaw and common tool binaries", () => {
       const content = requirePresetContent(policies.loadPreset("local-inference"));
       expect(content).toContain("/usr/local/bin/openclaw");