diff --git a/HISTORY.rst b/HISTORY.rst
index e1f5c31..199f22a 100644
--- a/HISTORY.rst
+++ b/HISTORY.rst
@@ -8,6 +8,10 @@ Changelog
 .. This document is user facing. Please word the changes in such a way
 .. that users understand how the changes affect the new version.
 
+unreleased
+---------------------------
++ ``-s`` pytest flag is now supported, streaming stdout/stderr as the tests run
+
 version 2.1.0
 ---------------------------
 + Python version 3.7 support is dropped because it is deprecated. Python
diff --git a/src/pytest_workflow/content_tests.py b/src/pytest_workflow/content_tests.py
index 2770188..51ba826 100644
--- a/src/pytest_workflow/content_tests.py
+++ b/src/pytest_workflow/content_tests.py
@@ -72,7 +72,7 @@ def check_content(strings: Iterable[str],
 
 class ContentTestCollector(pytest.Collector):
     def __init__(self, name: str, parent: pytest.Collector,
-                 filepath: Path,
+                 filepath: Optional[Path],
                  content_test: ContentTest,
                  workflow: Workflow,
                  content_name: Optional[str] = None):
@@ -105,6 +105,11 @@ def find_strings(self):
         When a file we test is not produced, we save the FileNotFoundError so
         we can give an accurate repr_failure."""
         self.workflow.wait()
+
+        if self.filepath is None:
+            self.file_not_found = True
+            return
+
         strings_to_check = (self.content_test.contains +
                             self.content_test.must_not_contain)
         patterns_to_check = (self.content_test.contains_regex +
@@ -195,6 +200,10 @@ def __init__(self, parent: ContentTestCollector, string: str,
         name = f"{contain} '{string}'"
         super().__init__(name, parent=parent)
         self.parent: ContentTestCollector = parent  # explicitly declare type
+        assert self.parent.filepath is not None, (
+            f"Invalid test {content_name}, unknown file to validate. "
+            "This can happen if you specify stdout/stderr tests while "
+            "specifying a different capture method.")
         self.should_contain = should_contain
         self.string = string
         self.content_name = content_name
diff --git a/src/pytest_workflow/plugin.py b/src/pytest_workflow/plugin.py
index 142c715..0b10c4a 100644
--- a/src/pytest_workflow/plugin.py
+++ b/src/pytest_workflow/plugin.py
@@ -409,7 +409,8 @@ def queue_workflow(self):
         workflow = Workflow(command=self.workflow_test.command,
                             cwd=tempdir,
                             name=self.workflow_test.name,
-                            desired_exit_code=self.workflow_test.exit_code)
+                            desired_exit_code=self.workflow_test.exit_code,
+                            capture=self.config.getoption("capture"))
 
         # Add the workflow to the workflow queue.
         self.config.workflow_queue.put(workflow)
diff --git a/src/pytest_workflow/workflow.py b/src/pytest_workflow/workflow.py
index 748b024..eec34f7 100644
--- a/src/pytest_workflow/workflow.py
+++ b/src/pytest_workflow/workflow.py
@@ -27,7 +27,11 @@
 import threading
 import time
 from pathlib import Path
-from typing import List, Optional
+from typing import List, Literal, Optional
+
+
+# pytest does not export this type
+CaptureMethod = Literal["fd", "sys", "no", "tee-sys"]
 
 
 class Workflow(object):
@@ -36,7 +40,8 @@ def __init__(self,
                  command: str,
                  cwd: Optional[Path] = None,
                  name: Optional[str] = None,
-                 desired_exit_code: int = 0):
+                 desired_exit_code: int = 0,
+                 capture: CaptureMethod = "fd"):
         """
         Initiates a workflow object
         :param command: The string that represents the command to be run
@@ -48,24 +53,35 @@ def __init__(self,
         if command == "":
             raise ValueError("command can not be an empty string")
         self.command = command
+
         # Always ensure a name. command.split()[0] can't fail because we tested
         # for emptiness.
         self.name = name or command.split()[0]
         self.cwd = cwd or Path()
+
         # For long running workflows it is best to save the stdout and stderr
         # to a file which can be checked with ``tail -f``.
         # stdout and stderr will be written to a tempfile if no CWD is given
         # to prevent clutter created when testing.
-        self.stdout_file = (
-            Path(tempfile.NamedTemporaryFile(prefix=self.name,
-                                             suffix=".out").name)
-            if cwd is None
-            else self.cwd / Path("log.out"))
-        self.stderr_file = (
-            Path(tempfile.NamedTemporaryFile(prefix=self.name,
-                                             suffix=".err").name)
-            if cwd is None
-            else self.cwd / Path("log.err"))
+        supported_capture_methods = ["no", "fd"]
+        if capture not in supported_capture_methods:
+            raise ValueError("only capture methods "
+                             f"{supported_capture_methods} are supported, "
+                             f"found {capture}")
+        self.capture = capture
+        self.stdout_file = None
+        self.stderr_file = None
+        if self.capture != "no":
+            self.stdout_file = (
+                Path(tempfile.NamedTemporaryFile(prefix=self.name,
+                                                 suffix=".out").name)
+                if cwd is None
+                else self.cwd / Path("log.out"))
+            self.stderr_file = (
+                Path(tempfile.NamedTemporaryFile(prefix=self.name,
+                                                 suffix=".err").name)
+                if cwd is None
+                else self.cwd / Path("log.err"))
         self._popen: Optional[subprocess.Popen] = None
         self._started = False
         self.errors: List[Exception] = []
@@ -79,9 +95,12 @@ def start(self):
         # is started from multiple threads.
         with self.start_lock:
             if not self._started:
+                stdout_h = None
+                stderr_h = None
                 try:
-                    stdout_h = self.stdout_file.open('wb')
-                    stderr_h = self.stderr_file.open('wb')
+                    if self.capture != "no":
+                        stdout_h = self.stdout_file.open('wb')
+                        stderr_h = self.stderr_file.open('wb')
                     sub_process_args = shlex.split(self.command)
                     self._popen = subprocess.Popen(
                         sub_process_args, stdout=stdout_h,
@@ -91,8 +110,10 @@ def start(self):
                     self.errors.append(error)
                 finally:
                     self._started = True
-                    stdout_h.close()
-                    stderr_h.close()
+                    if stdout_h is not None:
+                        stdout_h.close()
+                    if stderr_h is not None:
+                        stderr_h.close()
             else:
                 raise ValueError("Workflows can only be started once")
 
@@ -148,12 +169,20 @@ def matching_exitcode(self) -> bool:
     @property
     def stdout(self) -> bytes:
         self.wait()
+        if self.stdout_file is None:
+            raise ValueError(
+                f"Stdout not available with capture={self.capture}"
+            )
         with self.stdout_file.open('rb') as stdout:
             return stdout.read()
 
     @property
     def stderr(self) -> bytes:
         self.wait()
+        if self.stderr_file is None:
+            raise ValueError(
+                f"Stdout not available with capture={self.capture}"
+            )
         with self.stderr_file.open('rb') as stderr:
             return stderr.read()
 
diff --git a/tests/test_fail_messages.py b/tests/test_fail_messages.py
index 4640bd0..0e1cb35 100644
--- a/tests/test_fail_messages.py
+++ b/tests/test_fail_messages.py
@@ -179,3 +179,19 @@ def test_messages_exitcode(test: str, message: str, pytester):
     # possible due to multiple levels of process launching.
     result = pytester.runpytest("-v", "--sb", "5")
     assert message in result.stdout.str()
+
+
+def test_invalid_test_capture(pytester):
+    test_yml_contents = """\
+    - name: tee test
+      command: bash -c "echo foo"
+      stdout:
+        contains:
+          - foo
+    """
+    pytester.makefile(".yml", textwrap.dedent(test_yml_contents))
+    result = pytester.runpytest("-v", "-s")
+    assert (
+        "Invalid test 'tee test': stdout, unknown file to validate"
+        in result.stdout.str()
+    )
diff --git a/tests/test_workflow.py b/tests/test_workflow.py
index 4634427..3ef87a7 100644
--- a/tests/test_workflow.py
+++ b/tests/test_workflow.py
@@ -105,3 +105,17 @@ def test_workflow_matching_exit_code():
     workflow2 = Workflow("grep", desired_exit_code=2)
     workflow2.run()
     assert workflow2.matching_exitcode()
+
+
+def test_capture_unsupported():
+    with pytest.raises(ValueError) as error:
+        Workflow("echo moo", capture="tee-sys")
+    error.match("only capture methods")
+
+
+def test_capture_no():
+    workflow = Workflow("echo moo", capture="no")
+    workflow.run()
+    with pytest.raises(ValueError) as error:
+        workflow.stdout
+    error.match("Stdout not available")