Python: Allow optional MCP task tools through plain calls

docs/decisions/0045-python-mcp-v2-client-lifecycle.md

@@ -182,8 +182,9 @@ protocol-gated compatibility operations; they are not universal preflights.
 ### Tasks
 
 The supported MCP Python SDK 2.x runtime does not expose the current `io.modelcontextprotocol/tasks` extension.
-Tools that may require Tasks are not presented as ordinary supported calls, and the removed long-running Tasks
-sample remains deferred. An MRTR `InputRequiredResult` is not a Tasks `input_required` status.
+Tools that require Tasks are not presented as ordinary supported calls. Tools with optional Tasks support continue
+through ordinary `tools/call` without task augmentation, and the removed long-running Tasks sample remains deferred.
+An MRTR `InputRequiredResult` is not a Tasks `input_required` status.
 
 ## Consequences
 

python/packages/core/AGENTS.md

@@ -241,7 +241,7 @@ The vector store API is experimental under the shared `VECTOR_STORES` feature ID
   `cacheScope` hints. Legacy peers default to zero TTL, caller-supplied `ClientSession` connections stay uncached,
   and reconnect or effective header-identity changes replace a framework-owned Client's per-Client cache. Agent
   Framework does not expose shared-store or cache-mode configuration.
-- **MCP Tasks extension** - MCP Python SDK 2.2 does not implement the 2026 `io.modelcontextprotocol/tasks` runtime. Tools that advertise `execution.taskSupport == "required"` are skipped with a warning; optional and task-free tools continue through ordinary `tools/call`. Restoration of the experimental Tasks API and removed sample is tracked in microsoft/agent-framework#8245.
+- **MCP Tasks extension** - The current Agent Framework MCP client does not implement the `io.modelcontextprotocol/tasks` runtime. Tools that advertise `execution.taskSupport == "required"` are skipped with a warning; optional and task-free tools continue through ordinary `tools/call`. Tasks client support is tracked in microsoft/agent-framework#9215.
 
 ### File Access Harness (`_harness/_file_access.py`)
 

python/packages/core/agent_framework/_mcp.py

@@ -3112,10 +3112,10 @@ async def _load_tools_locked(self) -> None:
 
             for tool in tool_list.tools:
                 task_support = getattr(getattr(tool, "execution", None), "task_support", None)
-                if task_support in ["required", "optional"]:
+                if task_support == "required":
                     logger.warning(
-                        "Skipping MCP tool %r because it (might) require the Tasks extension, "
-                        "which MCP Python SDK 2.2 does not implement.",
+                        "Skipping MCP tool %r because it requires the Tasks extension, "
+                        "which this client does not support.",
                         tool.name,
                     )
                     continue

python/packages/core/tests/core/test_mcp.py

@@ -558,7 +558,7 @@ async def list_tools(params: types.PaginatedRequestParams | None = None) -> type
     assert tool._tool_param_names_by_name == ({} if empty_snapshot else {"keep": {"query"}})
 
 
-async def test_load_tools_skips_task_capable_tools(caplog: pytest.LogCaptureFixture) -> None:
+async def test_load_tools_skips_only_task_required_tools(caplog: pytest.LogCaptureFixture) -> None:
     tool = MCPTool(name="docs")  # type: ignore[abstract]  # ty: ignore[call-non-callable]
     mock_session = AsyncMock()
     tool.session = mock_session
@@ -593,19 +593,18 @@ async def test_load_tools_skips_task_capable_tools(caplog: pytest.LogCaptureFixt
     with caplog.at_level(logging.WARNING, logger=logger.name):
         await tool.load_tools()
 
-    assert [function.name for function in tool.functions] == ["forbidden", "ordinary"]
-    assert tool._tool_call_meta_by_name == {}
-    assert set(tool._tool_param_names_by_name) == {"forbidden", "ordinary"}
-    for tool_name in ("required", "optional"):
-        assert (
-            f"Skipping MCP tool '{tool_name}' because it (might) require the Tasks extension, "
-            "which MCP Python SDK 2.2 does not implement."
-        ) in caplog.messages
+    assert [function.name for function in tool.functions] == ["optional", "forbidden", "ordinary"]
+    assert tool._tool_call_meta_by_name == {"optional": {"mode": "optional"}}
+    assert set(tool._tool_param_names_by_name) == {"optional", "forbidden", "ordinary"}
+    skip_messages = [message for message in caplog.messages if message.startswith("Skipping MCP tool")]
+    assert skip_messages == [
+        "Skipping MCP tool 'required' because it requires the Tasks extension, which this client does not support."
+    ]
 
     for function in tool.functions:
         await function.invoke(arguments={})
 
-    assert [call.args[0] for call in mock_session.call_tool.await_args_list] == ["forbidden", "ordinary"]
+    assert [call.args[0] for call in mock_session.call_tool.await_args_list] == ["optional", "forbidden", "ordinary"]
 
 
 async def test_tool_refresh_preserves_prompt_when_server_advertises_same_raw_name() -> None: