diff --git a/ai-hub/app/config.py b/ai-hub/app/config.py
index 3e1b1e2..7f4668b 100644
--- a/ai-hub/app/config.py
+++ b/ai-hub/app/config.py
@@ -167,7 +167,10 @@
 
         if self.DB_MODE == "sqlite":
             normalized_path = local_db_path.lstrip("./")
-            self.DATABASE_URL: str = f"sqlite:///./{normalized_path}" if normalized_path else "sqlite:///./data/ai_hub.db"
+            if external_db_url and external_db_url.startswith("sqlite"):
+                self.DATABASE_URL: str = external_db_url
+            else:
+                self.DATABASE_URL: str = f"sqlite:///./{normalized_path}" if normalized_path else "sqlite:///./data/ai_hub.db"
         else:
             self.DATABASE_URL: str = external_db_url or "sqlite:///./data/ai_hub.db" 
         
diff --git a/ai-hub/integration_tests/conftest.py b/ai-hub/integration_tests/conftest.py
index a715d04..4b33842 100644
--- a/ai-hub/integration_tests/conftest.py
+++ b/ai-hub/integration_tests/conftest.py
@@ -122,42 +122,63 @@
                 "access_level": "use"
             })
 
-    # 5. Start Docker Containers for Nodes
-    print("[conftest] Starting local docker node containers...")
-    network = "cortex-hub_default"
+    # 5. Start Node Processes
+    is_docker_disabled = os.getenv("SKIP_DOCKER_NODES", "true").lower() == "true"
+    node_processes = []
     
-    # We must dynamically detect network if needed, but cortex-hub_default is expected.
-    # Kill any existing ones
-    subprocess.run(["docker", "rm", "-f", NODE_1, NODE_2], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
-    
-    # Get node image
-    image_proc = subprocess.run(["docker", "build", "-q", "./agent-node"], capture_output=True, text=True)
-    image_id = image_proc.stdout.strip()
-
-    for node_id in [NODE_1, NODE_2]:
-        cmd = [
-            "docker", "run", "-d",
-            "--name", node_id,
-            "--network", network,
-            "-e", f"AGENT_NODE_ID={node_id}",
-            "-e", f"AGENT_AUTH_TOKEN={tokens[node_id]}",
-            "-e", "GRPC_ENDPOINT=ai-hub:50051",
-            "-e", "HUB_URL=http://ai-hub:8000",
-            "-e", "AGENT_TLS_ENABLED=false",
-            image_id
-        ]
-        subprocess.run(cmd, check=True)
+    if not is_docker_disabled:
+        print("[conftest] Starting local docker node containers...")
+        network = "cortex-hub_default"
+        subprocess.run(["docker", "rm", "-f", NODE_1, NODE_2], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
+        image_proc = subprocess.run(["docker", "build", "-q", "./agent-node"], capture_output=True, text=True)
+        image_id = image_proc.stdout.strip()
+        for node_id in [NODE_1, NODE_2]:
+            cmd = ["docker", "run", "-d", "--name", node_id, "--network", network, "-e", f"AGENT_NODE_ID={node_id}", "-e", f"AGENT_AUTH_TOKEN={tokens[node_id]}", "-e", "GRPC_ENDPOINT=ai-hub:50051", "-e", "HUB_URL=http://ai-hub:8000", "-e", "AGENT_TLS_ENABLED=false", image_id]
+            subprocess.run(cmd, check=True)
+    else:
+        print("[conftest] Starting nodes as local Python background processes...")
+        # Resolve URLs for local process tests (assuming hub runs on localhost outside docker)
+        grpc_ep = os.getenv("TEST_GRPC_ENDPOINT", "127.0.0.1:50051")
+        http_ep = os.getenv("TEST_HUB_URL", "http://127.0.0.1:8000")
+        
+        # Determine the agent node source directory
+        agent_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "..", "agent-node"))
+        if not os.path.exists(agent_dir):
+             agent_dir = os.path.abspath(os.path.join(os.path.dirname(__file__), "..", "agent-node")) # Fallback
+             
+        for node_id in [NODE_1, NODE_2]:
+            env = os.environ.copy()
+            env["AGENT_NODE_ID"] = node_id
+            env["AGENT_AUTH_TOKEN"] = tokens[node_id]
+            env["GRPC_ENDPOINT"] = grpc_ep
+            env["HUB_URL"] = http_ep
+            env["AGENT_TLS_ENABLED"] = "false"
+            env["PYTHONPATH"] = f"{agent_dir}/src"
+            
+            proc = subprocess.Popen(
+                ["python3", "-m", "agent_node.node"],
+                env=env,
+                cwd=agent_dir,
+                stdout=subprocess.DEVNULL,
+                stderr=subprocess.DEVNULL
+            )
+            node_processes.append(proc)
 
     print("[conftest] Waiting for nodes to connect to mesh...")
-    time.sleep(5) # Let them handshake
+    time.sleep(5)
 
     client.close()
 
-    yield  # Run the tests!
+    yield node_processes
 
     # 6. Teardown
-    print("\n[conftest] Tearing down node containers...")
-    subprocess.run(["docker", "rm", "-f", NODE_1, NODE_2], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
+    print("\n[conftest] Tearing down nodes...")
+    if not is_docker_disabled:
+        subprocess.run(["docker", "rm", "-f", NODE_1, NODE_2], stdout=subprocess.DEVNULL, stderr=subprocess.DEVNULL)
+    else:
+        for proc in node_processes:
+            proc.terminate()
+            proc.wait()
 
 @pytest.fixture(autouse=True)
 def run_around_tests(setup_mesh_environment):
diff --git a/ai-hub/integration_tests/test_browser_llm.py b/ai-hub/integration_tests/test_browser_llm.py
index 5def577..608a36c 100644
--- a/ai-hub/integration_tests/test_browser_llm.py
+++ b/ai-hub/integration_tests/test_browser_llm.py
@@ -10,6 +10,7 @@
         "X-User-ID": os.environ.get("SYNC_TEST_USER_ID", "")
     }
 
+@pytest.mark.skipif(os.getenv("SKIP_DOCKER_NODES", "false").lower() == "true", reason="Browser skill requires a fully-loaded Docker container environment to access Chromium.")
 def test_browser_skill_weather():
     """
     Test explicitly asking the LLM context to leverage its browser skill 
diff --git a/run_integration_tests.sh b/run_integration_tests.sh
index 90e9718..d169837 100755
--- a/run_integration_tests.sh
+++ b/run_integration_tests.sh
@@ -27,43 +27,71 @@
     NO_REBUILD=true
 fi
 
-# Check if services are already running
-IS_RUNNING=false
-# We check if specifically our ai-hub containers are "Up" via compose
-if docker compose ps | grep -q 'Up'; then
-    IS_RUNNING=true
+# Check if docker daemon is reachable (i.e., not inside DevContainer without DIND)
+DOCKER_AVAILABLE=false
+if docker info >/dev/null 2>&1; then
+    DOCKER_AVAILABLE=true
+else
+    echo "Docker daemon not reachable (likely running in a Dev Container). Switching to Native Python mode..."
+    export SKIP_DOCKER_NODES=true
+    export SYNC_TEST_BASE_URL="http://127.0.0.1:8000/api/v1"
 fi
 
 if [ "$NO_REBUILD" = true ] && [ "$IS_RUNNING" = true ]; then
     echo "Service is already running and --no-rebuild flag provided."
     echo "Skipping rebuild and starting tests directly..."
 else
-    # 2. Clean start: purge the database / volumes
-    echo "Purging database and old containers..."
-    docker compose down -v --remove-orphans
-    docker kill test-node-1 test-node-2 2>/dev/null || true
-    docker rm test-node-1 test-node-2 2>/dev/null || true
+    if [ "$DOCKER_AVAILABLE" = true ]; then
+        # 2. Clean start: purge the database / volumes
+        echo "Purging database and old containers..."
+        docker compose down -v --remove-orphans
+        docker kill test-node-1 test-node-2 2>/dev/null || true
+        docker rm test-node-1 test-node-2 2>/dev/null || true
 
-    # 3. Build & start the Hub stack
-    echo "Starting AI Hub mesh..."
-    # Ensure permissions are clean
-    mkdir -p data
-    
-    # Force rebuild and clean start if we are explicitly rebuilding
-    docker compose build ai-hub ai-frontend
-    docker compose up -d
+        # 3. Build & start the Hub stack
+        echo "Starting AI Hub mesh..."
+        mkdir -p data
+        docker compose build ai-hub ai-frontend
+        docker compose up -d
 
-    # Wait for healthy
-    echo "Waiting for AI Hub to be ready..."
-    sleep 5
-    until curl -I -s http://localhost:8002/api/v1/users/login/local | grep -q "405"; do
-        echo "Waiting for AI Hub Backend..."
+        # Wait for healthy
+        echo "Waiting for AI Hub to be ready..."
+        sleep 5
+        until curl -I -s http://localhost:8002/api/v1/users/login/local | grep -q "405"; do
+            echo "Waiting for AI Hub Backend..."
+            sleep 2
+        done
+        sleep 3
+        echo "AI Hub Backend is online."
+    else
+        # Start AI Hub Backend natively via uvicorn
+        echo "Starting AI Hub natively in the background..."
+        export DATABASE_URL="sqlite:///./data/ai-hub-test.db"
+        export ENVIRONMENT="development"
+        export PATH_PREFIX="/api/v1"
+        # Purge local test database
+        rm -f ai-hub/data/ai-hub-test.db
+        mkdir -p ai-hub/data
+        
+        pkill -f uvicorn || true
+        pkill -f agent_node.node || true
+        sleep 1
+        
+        cd /app/ai-hub
+        uvicorn app.main:app --host 0.0.0.0 --port 8000 > native_hub.log 2>&1 &
+        HUB_PID=$!
+        cd /app
+        
+        # Wait for healthy
+        echo "Waiting for AI Hub to be ready..."
         sleep 2
-    done
-
-    # Wait for DB to fully initialize its tables
-    sleep 3
-    echo "AI Hub Backend is online."
+        until curl -I -s http://localhost:8000/api/v1/users/login/local | grep -q "405"; do
+            echo "Waiting for AI Hub Backend natively..."
+            sleep 2
+        done
+        sleep 3
+        echo "AI Hub Backend is online."
+    fi
 fi
 
 # 4. User setup via Pytest fixtures
@@ -78,7 +106,11 @@
     echo "=========================================="
     echo "    TEARING DOWN INTEGRATION ENVIRONMENT  "
     echo "=========================================="
-    docker compose down -v
+    if [ "$DOCKER_AVAILABLE" = true ]; then
+        docker compose down -v
+    else
+        kill $HUB_PID || true
+    fi
     echo "Done!"
 else
     echo "=========================================="