diff --git a/CHANGELOG.md b/CHANGELOG.md
index 1472de9..ad7dcd9 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -79,6 +79,9 @@ All notable changes to SyncMill are documented here.
 
 ### Fixed
 
+- Without a terminal, bare `syncmill start` and `start` without `--task`
+  print their refusal and the commands to use instead on separate lines. They
+  used to come out as one line joined by a literal `\x0a` (#254).
 - The TUI no longer exits when the board's decisions or the selected item
   cannot be read during a Board-tab refresh; it shows `DEGRADED` and keeps
   the last view, as it already did for the item list.
diff --git a/src/syncmill/cli.py b/src/syncmill/cli.py
index f455792..25e7884 100644
--- a/src/syncmill/cli.py
+++ b/src/syncmill/cli.py
@@ -976,8 +976,8 @@ def _start_project(
 
     if not task:
         if not interactive:
-            raise click.ClickException(
-                "start needs --task without a terminal.\n" + _PROJECT_STEPS
+            raise _MultilineError(
+                "start needs --task without a terminal.", *_PROJECT_STEPS.splitlines()
             )
         task = click.prompt("What should the agent do?").strip()
         if not task:
@@ -1228,10 +1228,10 @@ def start(
     if not trial and not project_flags:
         if not interactive:
             # Never wait for an answer nobody can give.
-            raise click.ClickException(
+            raise _MultilineError(
                 "start needs a terminal to ask what you want to do. Use "
-                "`syncmill start --trial` for the offline walkthrough.\n"
-                + _PROJECT_STEPS
+                "`syncmill start --trial` for the offline walkthrough.",
+                *_PROJECT_STEPS.splitlines(),
             )
         choice = click.prompt(
             "What would you like to do?\n"
diff --git a/tests/test_start.py b/tests/test_start.py
index 78dbe2a..dfd6c8c 100644
--- a/tests/test_start.py
+++ b/tests/test_start.py
@@ -95,9 +95,17 @@ def test_dry_run_shows_the_screen_and_runs_nothing(monkeypatch, project, calls):
     assert calls["run"] == []
 
 
+def _assert_project_steps_as_lines(output: str, first: str) -> None:
+    """The refusal and the steps it points at each sit on their own line: the
+    error boundary escapes a one-line message's LF to a visible ``\\x0a``."""
+    assert "\\x0a" not in output
+    assert output.splitlines() == [f"Error: {first}", *cli_module._PROJECT_STEPS.splitlines()]
+
+
 def test_without_a_terminal_it_needs_task_and_yes(monkeypatch, project, calls):
     no_task = _start(monkeypatch, ["-a", "codex"], ready=_ready("codex"))
     assert no_task.exit_code == 1 and "start needs --task" in no_task.output
+    _assert_project_steps_as_lines(no_task.output, "start needs --task without a terminal.")
 
     no_yes = _start(monkeypatch, ["--task", "t"], ready=_ready("codex"))
     assert no_yes.exit_code == 1 and "start needs --yes" in no_yes.output
@@ -109,6 +117,18 @@ def test_without_a_terminal_it_needs_task_and_yes(monkeypatch, project, calls):
     assert calls["diff"] == calls["apply"] == []  # nobody to ask; run printed the commands
 
 
+def test_bare_start_without_a_terminal_explains_instead_of_asking(monkeypatch, project, calls):
+    result = _start(monkeypatch, [], ready=_ready("codex"))
+
+    assert result.exit_code == 1
+    _assert_project_steps_as_lines(
+        result.output,
+        "start needs a terminal to ask what you want to do. Use "
+        "`syncmill start --trial` for the offline walkthrough.",
+    )
+    assert calls["run"] == []
+
+
 def test_no_ready_agent_points_at_the_trial(monkeypatch, project, calls):
     result = _start(
         monkeypatch, ["--task", "t"], ready=_ready(missing=["codex"], logged_out=["claude"])