diff --git a/tests/unit_tests/agent/README.md b/tests/unit_tests/agent/README.md
new file mode 100644
index 000000000..7e9ed0faf
--- /dev/null
+++ b/tests/unit_tests/agent/README.md
@@ -0,0 +1,154 @@
+# AWEL Agent Unit Tests
+
+This directory contains comprehensive unit tests for the DB-GPT AWEL (Agentic Workflow Expression Language) agent module.
+
+## Test Files
+
+### `test_agent_context.py`
+Tests for agent context classes from `dbgpt.agent.core.agent`:
+- **AgentContext**: Configuration and context for agent execution
+  - Initialization with default and custom values
+  - Dictionary serialization with `to_dict()`
+- **AgentGenerateContext**: Input context for agent generation
+  - Message and sender handling
+  - Rely messages and message chain processing
+  - Already_failed flag for error handling
+  - Dictionary serialization
+- **AgentMessage**: Message objects for agent communication
+  - Message creation, copying, and serialization
+  - Conversion to/from LLM message format
+  - Action report serialization
+  - Deep copy of context dictionaries
+- **AgentReviewInfo**: Review information for agent messages
+  - Approval and comments handling
+  - Serialization methods
+
+### `test_awel_agent_operator.py`
+Tests for AWEL agent operators from `dbgpt.agent.core.plan.awel.agent_operator`:
+- **AWELAgentOperator**: Main agent operator for AWEL workflows
+  - Normal execution flow with message passing
+  - Early return when `already_failed` is True
+  - Exception handling for empty messages/senders
+  - Fixed subgoal processing
+  - Message role conversion (HUMAN/AI)
+  - Failure scenario handling
+  - Agent building and configuration
+  - Memory cloning
+  - Begin agent matching and retry logic
+- **WrappedAgentOperator**: Agent wrapper for simple workflows
+  - Basic message passing and agent reply generation
+  - Empty message exception handling
+  - Current goal construction
+  - Memory handling
+- **AgentBranchOperator**: Branch operator for agent routing
+  - Branch map generation from downstream operators
+  - Next speakers routing logic
+  - Action report checking
+  - Filtering of non-AWEL operators
+
+## Running the Tests
+
+### Prerequisites
+The tests require the following packages to be installed:
+- pytest
+- pytest-asyncio
+- pytest-mock
+- dbgpt-core package and its dependencies
+
+### Installation
+```bash
+# Install dbgpt-core package in development mode
+cd packages/dbgpt-core
+pip install -e .
+
+# Or install test dependencies
+pip install pytest pytest-asyncio pytest-mock
+```
+
+### Running Tests
+```bash
+# Run all agent tests
+pytest tests/unit_tests/agent/ -v
+
+# Run specific test file
+pytest tests/unit_tests/agent/test_agent_context.py -v
+pytest tests/unit_tests/agent/test_awel_agent_operator.py -v
+
+# Run specific test class
+pytest tests/unit_tests/agent/test_agent_context.py::TestAgentContext -v
+
+# Run specific test method
+pytest tests/unit_tests/agent/test_agent_context.py::TestAgentContext::test_initialization_with_defaults -v
+
+# Run with coverage
+pytest tests/unit_tests/agent/ --cov=dbgpt.agent --cov-report=html
+```
+
+## Test Coverage
+
+The test suite covers:
+- ✅ Normal execution flows
+- ✅ Error handling and exceptions
+- ✅ Edge cases and boundary conditions
+- ✅ Message passing and transformations
+- ✅ Configuration and initialization
+- ✅ Serialization and deserialization
+- ✅ Mock-based isolation of external dependencies
+
+## Test Patterns
+
+### Using Fixtures
+Tests use pytest fixtures for common test data:
+```python
+@pytest.fixture
+def mock_agent():
+    """Create a mock Agent instance."""
+    agent = MagicMock(spec=Agent)
+    agent.name = "TestAgent"
+    return agent
+```
+
+### Async Testing
+Async tests use the `pytest.mark.asyncio` decorator:
+```python
+@pytest.mark.asyncio
+async def test_async_method(self, mock_agent):
+    result = await operator.map(context)
+    assert result is not None
+```
+
+### Parametrized Tests
+Common scenarios are tested with parametrization:
+```python
+@pytest.mark.parametrize(
+    "already_failed,should_execute",
+    [(True, False), (False, True)],
+)
+@pytest.mark.asyncio
+async def test_parametrized(already_failed, should_execute):
+    # Test implementation
+```
+
+## Mocking Strategy
+
+The tests use mocking to isolate external dependencies:
+- **Agent classes**: Mocked with `MagicMock(spec=Agent)`
+- **LLM clients**: Mocked to avoid actual API calls
+- **Resource managers**: Mocked to avoid database/file access
+- **Agent builders**: Mocked with builder pattern chains
+
+## Contributing
+
+When adding new tests:
+1. Follow the existing test structure and naming conventions
+2. Use descriptive docstrings for test methods
+3. Mock external dependencies appropriately
+4. Ensure tests are isolated and can run independently
+5. Format code with `ruff format tests/unit_tests/agent/`
+6. Check code style with `ruff check tests/unit_tests/agent/`
+
+## References
+
+- [pytest documentation](https://docs.pytest.org/)
+- [pytest-asyncio documentation](https://pytest-asyncio.readthedocs.io/)
+- [DB-GPT Agent Documentation](../../packages/dbgpt-core/src/dbgpt/agent/README.md)