diff --git a/codex-rs/app-server-protocol/schema/typescript/v2/TurnStartParams.ts b/codex-rs/app-server-protocol/schema/typescript/v2/TurnStartParams.ts
index ec390486410..d595035ddaf 100644
--- a/codex-rs/app-server-protocol/schema/typescript/v2/TurnStartParams.ts
+++ b/codex-rs/app-server-protocol/schema/typescript/v2/TurnStartParams.ts
@@ -37,5 +37,8 @@ personality?: Personality | null, /**
 outputSchema?: JsonValue | null, /**
  * EXPERIMENTAL - Set a pre-set collaboration mode.
  * Takes precedence over model, reasoning_effort, and developer instructions if set.
+ *
+ * For `collaboration_mode.settings.developer_instructions`, `null` means
+ * "use the built-in instructions for the selected mode".
  */
 collaborationMode?: CollaborationMode | null};
diff --git a/codex-rs/app-server-protocol/src/protocol/v2.rs b/codex-rs/app-server-protocol/src/protocol/v2.rs
index 630e0395786..738edd224ff 100644
--- a/codex-rs/app-server-protocol/src/protocol/v2.rs
+++ b/codex-rs/app-server-protocol/src/protocol/v2.rs
@@ -2069,6 +2069,9 @@ pub struct TurnStartParams {
 
     /// EXPERIMENTAL - Set a pre-set collaboration mode.
     /// Takes precedence over model, reasoning_effort, and developer instructions if set.
+    ///
+    /// For `collaboration_mode.settings.developer_instructions`, `null` means
+    /// "use the built-in instructions for the selected mode".
     #[experimental("turn/start.collaborationMode")]
     #[ts(optional = nullable)]
     pub collaboration_mode: Option<CollaborationMode>,
diff --git a/codex-rs/app-server/README.md b/codex-rs/app-server/README.md
index a6d49f2e4cd..6fc03fdd4b4 100644
--- a/codex-rs/app-server/README.md
+++ b/codex-rs/app-server/README.md
@@ -94,7 +94,7 @@ Example (from OpenAI's official VSCode extension):
 - `thread/unarchive` — move an archived rollout file back into the sessions directory; returns the restored `thread` on success.
 - `thread/compact/start` — trigger conversation history compaction for a thread; returns `{}` immediately while progress streams through standard turn/item notifications.
 - `thread/rollback` — drop the last N turns from the agent’s in-memory context and persist a rollback marker in the rollout so future resumes see the pruned history; returns the updated `thread` (with `turns` populated) on success.
-- `turn/start` — add user input to a thread and begin Codex generation; responds with the initial `turn` object and streams `turn/started`, `item/*`, and `turn/completed` notifications.
+- `turn/start` — add user input to a thread and begin Codex generation; responds with the initial `turn` object and streams `turn/started`, `item/*`, and `turn/completed` notifications. For `collaborationMode`, `settings.developer_instructions: null` means "use built-in instructions for the selected mode".
 - `turn/steer` — add user input to an already in-flight turn without starting a new turn; returns the active `turnId` that accepted the input.
 - `turn/interrupt` — request cancellation of an in-flight turn by `(thread_id, turn_id)`; success is an empty `{}` response and the turn finishes with `status: "interrupted"`.
 - `review/start` — kick off Codex’s automated reviewer for a thread; responds like `turn/start` and emits `item/started`/`item/completed` notifications with `enteredReviewMode` and `exitedReviewMode` items, plus a final assistant `agentMessage` containing the review.
diff --git a/codex-rs/app-server/src/codex_message_processor.rs b/codex-rs/app-server/src/codex_message_processor.rs
index f96daebc194..29fbd3d74d0 100644
--- a/codex-rs/app-server/src/codex_message_processor.rs
+++ b/codex-rs/app-server/src/codex_message_processor.rs
@@ -206,6 +206,7 @@ use codex_login::ServerOptions as LoginServerOptions;
 use codex_login::ShutdownHandle;
 use codex_login::run_login_server;
 use codex_protocol::ThreadId;
+use codex_protocol::config_types::CollaborationMode;
 use codex_protocol::config_types::ForcedLoginMethod;
 use codex_protocol::config_types::Personality;
 use codex_protocol::config_types::WindowsSandboxLevel;
@@ -397,6 +398,27 @@ impl CodexMessageProcessor {
             .unwrap_or_default()
     }
 
+    /// If a client sends `developer_instructions: null` during a mode switch,
+    /// use the built-in instructions for that mode.
+    fn normalize_turn_start_collaboration_mode(
+        &self,
+        mut collaboration_mode: CollaborationMode,
+    ) -> CollaborationMode {
+        if collaboration_mode.settings.developer_instructions.is_none()
+            && let Some(instructions) = self
+                .thread_manager
+                .list_collaboration_modes()
+                .into_iter()
+                .find(|preset| preset.mode == Some(collaboration_mode.mode))
+                .and_then(|preset| preset.developer_instructions.flatten())
+                .filter(|instructions| !instructions.is_empty())
+        {
+            collaboration_mode.settings.developer_instructions = Some(instructions);
+        }
+
+        collaboration_mode
+    }
+
     fn review_request_from_target(
         target: ApiReviewTarget,
     ) -> Result<(ReviewRequest, String), JSONRPCErrorError> {
@@ -4554,6 +4576,10 @@ impl CodexMessageProcessor {
             }
         };
 
+        let collaboration_mode = params
+            .collaboration_mode
+            .map(|mode| self.normalize_turn_start_collaboration_mode(mode));
+
         // Map v2 input items to core input items.
         let mapped_items: Vec<CoreInputItem> = params
             .input
@@ -4567,7 +4593,7 @@ impl CodexMessageProcessor {
             || params.model.is_some()
             || params.effort.is_some()
             || params.summary.is_some()
-            || params.collaboration_mode.is_some()
+            || collaboration_mode.is_some()
             || params.personality.is_some();
 
         // If any overrides are provided, update the session turn context first.
@@ -4581,7 +4607,7 @@ impl CodexMessageProcessor {
                     model: params.model,
                     effort: params.effort.map(Some),
                     summary: params.summary,
-                    collaboration_mode: params.collaboration_mode,
+                    collaboration_mode,
                     personality: params.personality,
                 })
                 .await;
diff --git a/codex-rs/app-server/tests/suite/v2/turn_start.rs b/codex-rs/app-server/tests/suite/v2/turn_start.rs
index 395561f69a1..27fc335cb44 100644
--- a/codex-rs/app-server/tests/suite/v2/turn_start.rs
+++ b/codex-rs/app-server/tests/suite/v2/turn_start.rs
@@ -408,6 +408,8 @@ async fn turn_start_accepts_collaboration_mode_override_v2() -> Result<()> {
     let request = response_mock.single_request();
     let payload = request.body_json();
     assert_eq!(payload["model"].as_str(), Some("mock-model-collab"));
+    let payload_text = payload.to_string();
+    assert!(payload_text.contains("The `request_user_input` tool is unavailable in Default mode."));
 
     Ok(())
 }
diff --git a/codex-rs/docs/codex_mcp_interface.md b/codex-rs/docs/codex_mcp_interface.md
index 8b1b5f70457..da415d9deb2 100644
--- a/codex-rs/docs/codex_mcp_interface.md
+++ b/codex-rs/docs/codex_mcp_interface.md
@@ -112,6 +112,8 @@ Fetch the built-in collaboration mode presets with `collaborationMode/list`. Thi
 - `data` – ordered list of collaboration mode masks (partial settings to apply on top of the base mode)
   - For tri-state fields like `reasoning_effort` and `developer_instructions`, omit the field to keep the current value, set it to `null` to clear it, or set a concrete value to update it.
 
+When sending `turn/start` with `collaborationMode`, `settings.developer_instructions: null` means "use built-in instructions for the selected mode".
+
 ## Event stream
 
 While a conversation runs, the server sends notifications: