diff --git a/.github/workflows/ci.yml b/.github/workflows/ci.yml
index 12338cb..09e7ef8 100644
--- a/.github/workflows/ci.yml
+++ b/.github/workflows/ci.yml
@@ -40,7 +40,20 @@ jobs:
           python-version: "3.11"
           cache: pip
       - run: pip install -r requirements.txt
-      - run: pytest tests/ -v --tb=short
+      # Coverage is scoped to production code: backend/.coveragerc omits
+      # tests/ so the gate reflects untested application code (with tests
+      # included the total was ~66%; production-only it is ~51%).
+      - run: |
+          pytest tests/ -v --tb=short \
+            --cov=. \
+            --cov-report=xml \
+            --cov-report=term-missing \
+            --cov-fail-under=50
+      - uses: actions/upload-artifact@v4
+        if: always()
+        with:
+          name: backend-coverage
+          path: backend/coverage.xml
 
   backend-security:
     name: Backend Security Scan
diff --git a/backend/.coveragerc b/backend/.coveragerc
new file mode 100644
index 0000000..a7cba45
--- /dev/null
+++ b/backend/.coveragerc
@@ -0,0 +1,3 @@
+[run]
+omit =
+    tests/*
diff --git a/backend/requirements.txt b/backend/requirements.txt
index 37850d1..4f260ed 100644
--- a/backend/requirements.txt
+++ b/backend/requirements.txt
@@ -26,6 +26,7 @@ nbformat>=5.10.0
 jupyter_client>=8.6.0
 pytest>=9.0.0
 pytest-asyncio>=1.3.0
+pytest-cov>=6.0.0
 httpx>=0.28.0
 
 # Paper + web search backends