diff --git a/docs/implementation-prompts/fix-multi-attach-node-affinity.md b/docs/implementation-prompts/fix-multi-attach-node-affinity.md
new file mode 100644
index 00000000..d4eaa549
--- /dev/null
+++ b/docs/implementation-prompts/fix-multi-attach-node-affinity.md
@@ -0,0 +1,79 @@
+# Implementation: Fix Multi-Attach PVC Node Affinity
+
+## Problem
+AgenticSessions get stuck in "Creating" phase when the temp-content initContainer and the Job's main pods are scheduled on different nodes while using ReadWriteOnce (RWO) PVCs. This causes a Multi-Attach error because RWO PVCs can only be mounted on one node at a time.
+
+## Root Cause
+In `components/operator/internal/handlers/sessions.go`, the Job spec creates:
+1. InitContainer `init-workspace` - prepares directories on the PVC
+2. InitContainer `temp-content` - starts content service temporarily
+3. Main containers `ambient-content` and `ambient-code-runner` - the actual workload
+
+When these containers are scheduled on different nodes, the PVC cannot attach to both nodes simultaneously.
+
+## Solution
+Add node affinity to ensure all containers in the Job run on the same node as the PVC attachment.
+
+## Code Changes Required
+
+**File**: `components/operator/internal/handlers/sessions.go`
+
+**Location**: In the `corev1.PodSpec` section (around line 385), add a node affinity configuration.
+
+**Change**: Add this field to the PodSpec struct after `AutomountServiceAccountToken`:
+
+```go
+Spec: corev1.PodSpec{
+    RestartPolicy: corev1.RestartPolicyNever,
+    AutomountServiceAccountToken: boolPtr(false),
+
+    // Add node affinity to ensure pod schedules on same node as PVC
+    Affinity: &corev1.Affinity{
+        PodAffinity: &corev1.PodAffinity{
+            PreferredDuringSchedulingIgnoredDuringExecution: []corev1.WeightedPodAffinityTerm{
+                {
+                    Weight: 100,
+                    PodAffinityTerm: corev1.PodAffinityTerm{
+                        LabelSelector: &v1.LabelSelector{
+                            MatchExpressions: []v1.LabelSelectorRequirement{
+                                {
+                                    Key:      "agentic-session",
+                                    Operator: v1.LabelSelectorOpIn,
+                                    Values:   []string{name},
+                                },
+                            },
+                        },
+                        TopologyKey: "kubernetes.io/hostname",
+                    },
+                },
+            },
+        },
+    },
+
+    Volumes: []corev1.Volume{
+        // ... existing volumes ...
+```
+
+## Testing
+
+1. Create a new AgenticSession
+2. Verify the session pod starts without Multi-Attach errors
+3. Check pod events: `oc describe pod <pod-name> -n <namespace>`
+4. Verify all containers are on the same node: `oc get pods -o wide`
+
+## Alternative Approaches Considered
+
+- **ReadWriteMany (RWX) PVC**: Would allow multi-node attachment but requires NFS or similar storage backend (not always available)
+- **Single InitContainer**: Would work but requires refactoring the workspace prep and content service startup logic
+- **Node Selector**: Too rigid, doesn't work well with cluster autoscaling
+
+The node affinity approach is preferred because it's least invasive and works with existing RWO storage.
+
+## Rollback
+
+If this causes issues, revert the commit and manually delete stuck pods:
+```bash
+git revert <commit-sha>
+make deploy NAMESPACE=<namespace>
+oc delete pod -l agentic-session=<session-name> -n <namespace>
+```