diff --git a/agents.md b/agents.md
index 3c06574..94a5d46 100644
--- a/agents.md
+++ b/agents.md
@@ -1,6 +1,5 @@
 # Agents Upgrade Plan
 
-- test: rebuild ollama integration suite to exercise local, cloud, search, and quota error flows
 - refactor: simplify provider/session coupling so compute_web_search_settings shares normalization utilities
 - docs: clarify cloud vs local env configuration (api key precedence, insecure toggle) in configuration guide
 - ci: add targeted regression job for ollama providers covering chat, streaming, and tool execution paths
diff --git a/crates/owlen-core/tests/ollama_wiremock.rs b/crates/owlen-core/tests/ollama_wiremock.rs
index a0e2d28..bdaeff6 100644
--- a/crates/owlen-core/tests/ollama_wiremock.rs
+++ b/crates/owlen-core/tests/ollama_wiremock.rs
@@ -113,7 +113,7 @@ fn configure_local(base_url: &str) -> Config {
     config
 }
 
-fn configure_cloud(base_url: &str, search_endpoint: &str) -> Config {
+fn configure_cloud(base_url: &str, search_endpoint: Option<&str>) -> Config {
     let mut config = Config::default();
     config.general.default_provider = "ollama_cloud".into();
     config.general.default_model = Some("llama3:8b-cloud".into());
@@ -130,10 +130,11 @@ fn configure_cloud(base_url: &str, search_endpoint: &str) -> Config {
         cloud.enabled = true;
         cloud.base_url = Some(base_url.to_string());
         cloud.api_key = Some("test-key".into());
-        cloud.extra.insert(
-            "web_search_endpoint".into(),
-            Value::String(search_endpoint.into()),
-        );
+        if let Some(endpoint) = search_endpoint {
+            cloud
+                .extra
+                .insert("web_search_endpoint".into(), Value::String(endpoint.into()));
+        }
         cloud.extra.insert(
             owlen_core::config::OLLAMA_CLOUD_ENDPOINT_KEY.into(),
             Value::String(base_url.to_string()),
@@ -180,14 +181,13 @@ async fn run_local_completion(base_suffix: &str) {
     assert_eq!(snapshot.weekly.total_tokens, 36);
 }
 
-async fn run_cloud_tool_flow(base_suffix: &str) {
+async fn run_cloud_tool_flow(
+    base_suffix: &str,
+    search_endpoint: Option<&str>,
+    expected_search_path: &str,
+) {
     let server = MockServer::start().await;
     let raw_base = format!("{}{}", server.uri(), base_suffix);
-    let search_endpoint = if base_suffix.is_empty() {
-        "/v1/web/search"
-    } else {
-        "/web/search"
-    };
 
     let tags = load_fixture("ollama_tags");
     let tool_call = load_fixture("ollama_cloud_tool_call");
@@ -219,7 +219,7 @@ async fn run_cloud_tool_flow(base_suffix: &str) {
         .await;
 
     Mock::given(method("POST"))
-        .and(path("/v1/web/search"))
+        .and(path(expected_search_path))
         .and(header("authorization", "Bearer test-key"))
         .respond_with(ResponseTemplate::new(200).set_body_json(json!({
             "results": [
@@ -254,7 +254,7 @@ async fn run_cloud_tool_flow(base_suffix: &str) {
     let search_url = format!(
         "{}/{}",
         raw_base.trim_end_matches('/'),
-        search_endpoint.trim_start_matches('/')
+        expected_search_path.trim_start_matches('/')
     );
 
     session.grant_consent(
@@ -325,7 +325,7 @@ async fn run_cloud_error(
         .mount(&server)
         .await;
 
-    let config = configure_cloud(&raw_base, search_endpoint);
+    let config = configure_cloud(&raw_base, Some(search_endpoint));
     let cloud_cfg = config
         .providers
         .get("ollama_cloud")
@@ -368,12 +368,17 @@ async fn local_provider_accepts_v1_base_url() {
 
 #[tokio::test(flavor = "multi_thread")]
 async fn cloud_tool_call_flows_through_web_search() {
-    run_cloud_tool_flow("").await;
+    run_cloud_tool_flow("", Some("/v1/web/search"), "/v1/web/search").await;
 }
 
 #[tokio::test(flavor = "multi_thread")]
 async fn cloud_tool_call_accepts_v1_base_url() {
-    run_cloud_tool_flow("/v1").await;
+    run_cloud_tool_flow("/v1", Some("/web/search"), "/v1/web/search").await;
+}
+
+#[tokio::test(flavor = "multi_thread")]
+async fn cloud_tool_call_uses_default_search_endpoint() {
+    run_cloud_tool_flow("", None, "/api/web_search").await;
 }
 
 #[tokio::test(flavor = "multi_thread")]