@@ -37,6 +37,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 - Added a `get_preferences` operation to the `cost-optimization-hub` tool (`GetPreferences`), returning the calling account's savings estimation mode, member account discount visibility, and preferred commitment (term and payment option).
 
 ### Changed
+- **Breaking:** `ri-performance` `get_reservation_coverage` returns only `on_demand_cost` in `coverage_cost`. The `reserved_cost`, `total_cost` and `coverage_cost_percentage` fields are removed: Cost Explorer's `CoverageCost` has only `OnDemandCost`, so they were always `null`. Use `coverage_hours` (or `coverage_normalized_units` for EC2) for coverage percentages.
 - **Breaking:** `sp-performance` now labels the request window `requested_time_period` instead of `time_period` on all three operations. A MONTHLY utilization entry's `TimePeriod` is labelled with its calendar month (a 2026-08-17 start returns a first entry labelled 2026-08-01, whose figures still start on the 17th), and the description now says so. `get_savings_plans_utilization_details` now returns the API's `TimePeriod` as `time_period` (the requested start, and an end the API may move earlier to the first day without complete data), and returns `total` only when the whole query came back in one page: the API's `Total` covers only the Savings Plans on its own page, so a multi-page, partial (`max_items`) or `next_token` continuation result previously reported one page's figure as the query's (live: page 1 of 3 reported 100% utilization against 99.71% for the query). `get_savings_plans_coverage` no longer reads a `Total` (the GetSavingsPlansCoverage API does not return one) and reports `granularity` only when it was sent, not with `group_by`. The description lists the utilization details filter dimensions and says which period to report.
 - **Breaking:** `ri-performance` grouped `get_reservation_coverage` and `get_reservation_utilization` return a period's `total` only when the whole query came back in one page. With `group_by`, the API's period `Total` sums only the groups on its own page, and the tool kept page 1's as the period's: live, a 3-group page of an 877-group coverage query reported 638,311 running hours against 902,438,217 for the query, and a `next_token` continuation reported its own page's 1,311,271. `total` is now omitted from a multi-page, partial (`max_items`/`max_pages`) or `next_token` continuation result, and the description points to the ungrouped call for the overall figure. A `total` the API did not return is omitted rather than returned as `{}`. Ungrouped results, which are always one page, are unchanged.
 - **Breaking:** every paginated tool operation now takes `max_items` (the total number of items to return) and `next_token`. The `max_results`, `max_pages` and `page_size` parameters are removed (`ri-performance` keeps `max_pages` for `get_reservation_utilization`, below), and `next_page_token` (`sp-purchase-analyzer`, `sp-recommendation`) is renamed to `next_token`. Without `max_items` an operation fetches every page; with it, the operation stops before a page that could take the result past `max_items` (the first page is always returned) and returns the service's own continuation token unchanged. `ri-performance` `get_reservation_utilization` grouped by `SUBSCRIPTION_ID` takes `max_pages` (pages of 500 groups) instead of `max_items`, because its token only works with the same page size; `max_items` there returns `validation_error`. Billing Conductor `get-billing-group-cost-report` requests pages of 200 to 300 results, the range the service accepts, so a `max_items` below 200 can return up to 200. One call stops paging, on a page boundary, after `BCM_MCP_PAGINATION_MAX_SECONDS` (default 300) seconds. Paginated responses carry a lowercase `pagination` block (`complete_dataset`, `has_more`, `next_token`, `items_returned`) ahead of the data, inline and in the SQL-offload result; `has_more` is true only when there is a `next_token` to continue with, and `complete_dataset` reports whether the result is partial; a partial result also leads with top-level `complete_dataset: false` and a `warning`. The old `Pagination` block, top-level `next_token`, `has_more_results`, `total_count`, `pages_fetched`, `total_results` and `duration_ms` fields are removed. A rejected or expired `next_token` returns `error_type: invalid_next_token` (`retry_strategy: modify_request`). Operations that are not paginated take neither parameter. See "Pagination" in the README.
 
@@ -569,14 +569,13 @@ def format_coverage_metrics(coverage_data: Dict) -> Dict:
             'coverage_normalized_units_percentage': cnu.get('CoverageNormalizedUnitsPercentage'),
         }
 
-    # Format cost coverage
+    # Format cost coverage. CoverageCost holds only OnDemandCost (the on-demand cost of
+    # usage reservations did not cover); Cost Explorer reports no reserved cost or cost
+    # coverage percentage, so use coverage_hours_percentage for coverage.
     if 'CoverageCost' in coverage_data:
         cc = coverage_data['CoverageCost']
         formatted_coverage['coverage_cost'] = {
             'on_demand_cost': cc.get('OnDemandCost'),
-            'reserved_cost': cc.get('ReservedCost'),
-            'total_cost': cc.get('TotalCost'),
-            'coverage_cost_percentage': cc.get('CoverageCostPercentage'),
         }
 
     return formatted_coverage
 
@@ -22,6 +22,7 @@
 - Error handling for invalid date ranges and missing reservation data
 """
 
+import botocore.session
 import pytest
 from awslabs.billing_cost_management_mcp_server.tools.ri_performance_tools import (
     format_coverage_metrics,
@@ -63,9 +64,6 @@ async def ri_performance(ctx, operation, **kwargs):
                     },
                     'coverage_cost': {
                         'on_demand_cost': '10.0',
-                        'reserved_cost': '30.0',
-                        'total_cost': '40.0',
-                        'coverage_cost_percentage': '75.0',
                     },
                 },
             },
@@ -118,9 +116,6 @@ def mock_ce_client():
                     },
                     'CoverageCost': {
                         'OnDemandCost': '10.0',
-                        'ReservedCost': '30.0',
-                        'TotalCost': '40.0',
-                        'CoverageCostPercentage': '75.0',
                     },
                 },
                 'Groups': [
@@ -138,9 +133,6 @@ def mock_ce_client():
                             },
                             'CoverageCost': {
                                 'OnDemandCost': '5.0',
-                                'ReservedCost': '15.0',
-                                'TotalCost': '20.0',
-                                'CoverageCostPercentage': '75.0',
                             },
                         },
                     },
@@ -158,9 +150,6 @@ def mock_ce_client():
                             },
                             'CoverageCost': {
                                 'OnDemandCost': '5.0',
-                                'ReservedCost': '15.0',
-                                'TotalCost': '20.0',
-                                'CoverageCostPercentage': '75.0',
                             },
                         },
                     },
@@ -176,9 +165,6 @@ def mock_ce_client():
             },
             'CoverageCost': {
                 'OnDemandCost': '10.0',
-                'ReservedCost': '30.0',
-                'TotalCost': '40.0',
-                'CoverageCostPercentage': '75.0',
             },
         },
     }
@@ -234,6 +220,49 @@ def mock_ce_client():
     return mock_client
 
 
+def _undeclared_members(value, shape, path='') -> list:
+    """Paths in a mocked response that the botocore output shape does not declare."""
+    if shape.type_name == 'structure' and isinstance(value, dict):
+        found = []
+        for key, item in value.items():
+            if key not in shape.members:
+                found.append(f'{path}.{key}')
+            else:
+                found += _undeclared_members(item, shape.members[key], f'{path}.{key}')
+        return found
+    if shape.type_name == 'list' and isinstance(value, list):
+        return [
+            p
+            for i, item in enumerate(value)
+            for p in _undeclared_members(item, shape.member, f'{path}[{i}]')
+        ]
+    if shape.type_name == 'map' and isinstance(value, dict):
+        return [
+            p
+            for k, item in value.items()
+            for p in _undeclared_members(item, shape.value, f'{path}[{k}]')
+        ]
+    return []
+
+
+@pytest.mark.parametrize(
+    'method,operation',
+    [
+        ('get_reservation_coverage', 'GetReservationCoverage'),
+        ('get_reservation_utilization', 'GetReservationUtilization'),
+    ],
+)
+def test_mock_ce_responses_only_use_botocore_output_members(mock_ce_client, method, operation):
+    """Mocked Cost Explorer responses hold only members the real API can return."""
+    output_shape = (
+        botocore.session.get_session()
+        .get_service_model('ce')
+        .operation_model(operation)
+        .output_shape
+    )
+    assert _undeclared_members(getattr(mock_ce_client, method).return_value, output_shape) == []
+
+
 def test_format_coverage_metrics():
     """Test format_coverage_metrics function."""
     # Setup
@@ -246,9 +275,6 @@ def test_format_coverage_metrics():
         },
         'CoverageCost': {
             'OnDemandCost': '10.0',
-            'ReservedCost': '30.0',
-            'TotalCost': '40.0',
-            'CoverageCostPercentage': '75.0',
         },
         'CoverageNormalizedUnits': {
             'OnDemandNormalizedUnits': '200.0',
@@ -270,9 +296,8 @@ def test_format_coverage_metrics():
     assert result['coverage_hours']['reserved_hours'] == '400.0'
     assert result['coverage_hours']['coverage_hours_percentage'] == '80.0'
 
-    assert result['coverage_cost']['on_demand_cost'] == '10.0'
-    assert result['coverage_cost']['reserved_cost'] == '30.0'
-    assert result['coverage_cost']['coverage_cost_percentage'] == '75.0'
+    # CoverageCost has only OnDemandCost; no always-null reserved/total/percentage keys.
+    assert result['coverage_cost'] == {'on_demand_cost': '10.0'}
 
     assert result['coverage_normalized_units']['on_demand_normalized_units'] == '200.0'
     assert result['coverage_normalized_units']['reserved_normalized_units'] == '800.0'