diff --git a/bin/lib/onboard.js b/bin/lib/onboard.js
index 23f19b01a89..e70a386b0a2 100644
--- a/bin/lib/onboard.js
+++ b/bin/lib/onboard.js
@@ -647,11 +647,13 @@ async function setupInference(sandboxName, model, provider) {
 
   if (provider === "nvidia-nim") {
     // Create nvidia-nim provider
+    // Use bare key form (--credential NVIDIA_API_KEY) so the actual key value
+    // is passed via the child process environment, not visible in `ps aux`.
     run(
       `openshell provider create --name nvidia-nim --type openai ` +
-      `--credential "NVIDIA_API_KEY=${process.env.NVIDIA_API_KEY}" ` +
+      `--credential NVIDIA_API_KEY ` +
       `--config "OPENAI_BASE_URL=https://integrate.api.nvidia.com/v1" 2>&1 || true`,
-      { ignoreError: true }
+      { ignoreError: true, env: { NVIDIA_API_KEY: process.env.NVIDIA_API_KEY } }
     );
     run(
       `openshell inference set --no-verify --provider nvidia-nim --model ${model} 2>/dev/null || true`,
@@ -666,11 +668,11 @@ async function setupInference(sandboxName, model, provider) {
     const baseUrl = getLocalProviderBaseUrl(provider);
     run(
       `openshell provider create --name vllm-local --type openai ` +
-      `--credential "OPENAI_API_KEY=dummy" ` +
+      `--credential OPENAI_API_KEY ` +
       `--config "OPENAI_BASE_URL=${baseUrl}" 2>&1 || ` +
-      `openshell provider update vllm-local --credential "OPENAI_API_KEY=dummy" ` +
+      `openshell provider update vllm-local --credential OPENAI_API_KEY ` +
       `--config "OPENAI_BASE_URL=${baseUrl}" 2>&1 || true`,
-      { ignoreError: true }
+      { ignoreError: true, env: { OPENAI_API_KEY: "dummy" } }
     );
     run(
       `openshell inference set --no-verify --provider vllm-local --model ${model} 2>/dev/null || true`,
@@ -686,11 +688,11 @@ async function setupInference(sandboxName, model, provider) {
     const baseUrl = getLocalProviderBaseUrl(provider);
     run(
       `openshell provider create --name ollama-local --type openai ` +
-      `--credential "OPENAI_API_KEY=ollama" ` +
+      `--credential OPENAI_API_KEY ` +
       `--config "OPENAI_BASE_URL=${baseUrl}" 2>&1 || ` +
-      `openshell provider update ollama-local --credential "OPENAI_API_KEY=ollama" ` +
+      `openshell provider update ollama-local --credential OPENAI_API_KEY ` +
       `--config "OPENAI_BASE_URL=${baseUrl}" 2>&1 || true`,
-      { ignoreError: true }
+      { ignoreError: true, env: { OPENAI_API_KEY: "ollama" } }
     );
     run(
       `openshell inference set --no-verify --provider ollama-local --model ${model} 2>/dev/null || true`,
diff --git a/nemoclaw-blueprint/orchestrator/runner.py b/nemoclaw-blueprint/orchestrator/runner.py
index 432c228c3be..4d5a5e3ac44 100644
--- a/nemoclaw-blueprint/orchestrator/runner.py
+++ b/nemoclaw-blueprint/orchestrator/runner.py
@@ -57,13 +57,18 @@ def run_cmd(
     *,
     check: bool = True,
     capture: bool = False,
+    extra_env: dict[str, str] | None = None,
 ) -> subprocess.CompletedProcess[str]:
     """Run a command as an argv list (never shell=True)."""
+    env = None
+    if extra_env:
+        env = {**os.environ, **extra_env}
     return subprocess.run(
         args,
         check=check,
         capture_output=capture,
         text=True,
+        env=env,
     )
 
 
@@ -209,12 +214,16 @@ def action_apply(
         "--type",
         provider_type,
     ]
-    if credential:
-        provider_args.extend(["--credential", f"OPENAI_API_KEY={credential}"])
+    # Use bare key form (--credential OPENAI_API_KEY) so the actual credential
+    # value is passed via the child process environment, not visible in `ps aux`.
+    provider_env: dict[str, str] = {}
+    if credential and credential_env:
+        provider_args.extend(["--credential", credential_env])
+        provider_env[credential_env] = credential
     if endpoint:
         provider_args.extend(["--config", f"OPENAI_BASE_URL={endpoint}"])
 
-    run_cmd(provider_args, check=False, capture=True)
+    run_cmd(provider_args, check=False, capture=True, extra_env=provider_env)
 
     # Step 3: Set inference route
     progress(70, "Setting inference route")
diff --git a/nemoclaw/src/commands/onboard.ts b/nemoclaw/src/commands/onboard.ts
index 72fb9fcdd4c..6843adabc31 100644
--- a/nemoclaw/src/commands/onboard.ts
+++ b/nemoclaw/src/commands/onboard.ts
@@ -216,10 +216,11 @@ async function promptEndpoint(
   return (await promptSelect("Select your inference endpoint:", options)) as EndpointType;
 }
 
-function execOpenShell(args: string[]): string {
+function execOpenShell(args: string[], options?: { env?: Record<string, string> }): string {
   return execFileSync("openshell", args, {
     encoding: "utf-8",
     stdio: ["pipe", "pipe", "pipe"],
+    env: { ...process.env, ...options?.env },
   });
 }
 
@@ -450,10 +451,10 @@ export async function cliOnboard(opts: OnboardOptions): Promise<void> {
       "--type",
       "openai",
       "--credential",
-      `${credentialEnv}=${apiKey}`,
+      credentialEnv,
       "--config",
       `OPENAI_BASE_URL=${endpointUrl}`,
-    ]);
+    ], { env: { [credentialEnv]: apiKey } });
     logger.info(`Created provider: ${providerName}`);
   } catch (err) {
     const stderr =
@@ -465,10 +466,10 @@ export async function cliOnboard(opts: OnboardOptions): Promise<void> {
           "update",
           providerName,
           "--credential",
-          `${credentialEnv}=${apiKey}`,
+          credentialEnv,
           "--config",
           `OPENAI_BASE_URL=${endpointUrl}`,
-        ]);
+        ], { env: { [credentialEnv]: apiKey } });
         logger.info(`Updated provider: ${providerName}`);
       } catch (updateErr) {
         const updateStderr =