diff --git a/codex-rs/app-server-protocol/src/protocol.rs b/codex-rs/app-server-protocol/src/protocol.rs
index 98b16a26e5..74e9b6f9e2 100644
--- a/codex-rs/app-server-protocol/src/protocol.rs
+++ b/codex-rs/app-server-protocol/src/protocol.rs
@@ -266,6 +266,10 @@ pub struct NewConversationParams {
     #[serde(skip_serializing_if = "Option::is_none")]
     pub model: Option<String>,
 
+    /// Override the model provider to use for this session.
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub model_provider: Option<String>,
+
     /// Configuration profile from config.toml to specify default options.
     #[serde(skip_serializing_if = "Option::is_none")]
     pub profile: Option<String>,
@@ -1032,6 +1036,7 @@ mod tests {
             request_id: RequestId::Integer(42),
             params: NewConversationParams {
                 model: Some("gpt-5-codex".to_string()),
+                model_provider: None,
                 profile: None,
                 cwd: None,
                 approval_policy: Some(AskForApproval::OnRequest),
diff --git a/codex-rs/app-server/src/codex_message_processor.rs b/codex-rs/app-server/src/codex_message_processor.rs
index f6f81173b2..9b66322047 100644
--- a/codex-rs/app-server/src/codex_message_processor.rs
+++ b/codex-rs/app-server/src/codex_message_processor.rs
@@ -1610,6 +1610,7 @@ async fn derive_config_from_params(
 ) -> std::io::Result<Config> {
     let NewConversationParams {
         model,
+        model_provider,
         profile,
         cwd,
         approval_policy,
@@ -1625,7 +1626,7 @@ async fn derive_config_from_params(
         cwd: cwd.map(PathBuf::from),
         approval_policy,
         sandbox_mode,
-        model_provider: None,
+        model_provider,
         codex_linux_sandbox_exe,
         base_instructions,
         include_apply_patch_tool,