feat(tui): an Evaluate button beside View Traces and Deploy

A run finishes in the terminal UI and the next question is usually "was that
any good?" — which `crewai eval` answers, from the command line, about the run
that just happened. The button is there now, with `e` bound to it.

It behaves like Deploy rather than like View Traces: it leaves the UI and then
runs the command, because `crewai eval` talks — it prints the link to follow,
waits for the verdict and prints that — and none of it belongs inside a
full-screen app. It calls `eval_crew()` itself rather than reimplementing any
of it, so the button and the command can never say different things; a clean
SystemExit is the command deciding, and a failure is reported without failing
the run that already succeeded.

Declarative flows get it too — same button, same tail.

Co-Authored-By: Claude Opus 5 (1M context) <[email protected]>

lib/cli/src/crewai_cli/crew_run_tui.py

@@ -421,6 +421,19 @@ FooterKey .footer-key--key {
     color: #888888;
 }
 
+#btn-eval {
+    background: #2b2b2b;
+    color: #e0e0e0;
+    border: none;
+}
+#btn-eval:hover {
+    background: #3d3d3d;
+}
+#btn-eval:disabled {
+    background: #202020;
+    color: #888888;
+}
+
 #btn-deploy {
     background: #333333;
     color: #e0e0e0;
@@ -437,6 +450,7 @@ FooterKey .footer-key--key {
         Binding("s", "toggle_sidebar", "Sidebar"),
         Binding("l", "toggle_logs", "Logs"),
         Binding("t", "view_traces", "Traces", show=False),
+        Binding("e", "evaluate_crew", "Evaluate", show=False),
         Binding("d", "deploy_crew", "Deploy", show=False),
         Binding("down", "log_down", "Log ↓", show=False),
         Binding("up", "log_up", "Log ↑", show=False),
@@ -529,6 +543,7 @@ FooterKey .footer-key--key {
         self._current_method: str | None = None
         self._elapsed_frozen: float | None = None
         self._want_deploy: bool = False
+        self._want_eval: bool = False
         self._consent_screen: TraceConsentScreen | None = None
         self._trace_consent_pending: threading.Event | None = None
         self._discard_trace_on_exit = False
@@ -557,6 +572,7 @@ FooterKey .footer-key--key {
                 yield Static(id="sidebar-content")
                 with Vertical(id="sidebar-actions"):
                     yield Button("View Traces", id="btn-traces", classes="action-btn")
+                    yield Button("Evaluate", id="btn-eval", classes="action-btn")
                     yield Button("Deploy", id="btn-deploy", classes="action-btn")
             with Vertical(id="main-panel"):
                 yield Static(id="task-header")
@@ -1066,6 +1082,21 @@ FooterKey .footer-key--key {
         self._unsubscribe()
         self.exit(self._crew_result)
 
+    def action_evaluate_crew(self) -> None:
+        """Grade the run that just finished.
+
+        Like Deploy, this leaves the terminal UI first: `crewai eval` talks —
+        it prints the link to follow, waits for the verdict and prints it — and
+        that belongs in the terminal, not inside a full-screen app.
+        """
+        if self._status != "completed":
+            return
+
+        self._record_tui_button_click("evaluate")
+        self._want_eval = True
+        self._unsubscribe()
+        self.exit(self._crew_result)
+
     def _record_tui_button_click(self, button_name: str) -> None:
         try:
             if self._telemetry is None:
@@ -1078,6 +1109,8 @@ FooterKey .footer-key--key {
     def on_button_pressed(self, event: Button.Pressed) -> None:
         if event.button.id == "btn-traces":
             self.action_view_traces()
+        elif event.button.id == "btn-eval":
+            self.action_evaluate_crew()
         elif event.button.id == "btn-deploy":
             self.action_deploy_crew()
 

lib/cli/src/crewai_cli/run_crew.py

@@ -370,6 +370,9 @@ def _run_json_crew(
     if getattr(app, "_want_deploy", False):
         _chain_deploy()
 
+    if getattr(app, "_want_eval", False):
+        _chain_eval()
+
     return app._crew_result
 
 
@@ -493,6 +496,29 @@ def _run_json_crew_in_project_env(
     return None
 
 
+def _chain_eval() -> None:
+    """Grade the run that just finished, in the terminal the TUI has left.
+
+    `crewai eval` is the whole command: it finds the run crewAI recorded, asks
+    AMP to evaluate it, prints the link to follow and then the verdict. Running
+    it here rather than reimplementing it means the button and the command can
+    never say different things.
+    """
+    from rich.console import Console
+
+    console = Console()
+    try:
+        from crewai_cli.experimental.eval_crew import eval_crew
+
+        console.print("\nEvaluating this run…\n", style="bold #FF5A50")
+        eval_crew()
+    except SystemExit as exc:  # the command says why and picks the code
+        if isinstance(exc.code, int) and exc.code not in (0, None):
+            raise
+    except Exception as error:  # noqa: BLE001 - a failed evaluation never fails the run
+        console.print(f"\nEvaluation failed: {error}\n", style="bold red")
+
+
 def _chain_deploy() -> None:
     from rich.console import Console
 

lib/cli/src/crewai_cli/run_declarative_flow.py

@@ -218,6 +218,11 @@ def _run_declarative_flow_tui(
 
         _chain_deploy()
 
+    if getattr(app, "_want_eval", False):
+        from crewai_cli.run_crew import _chain_eval
+
+        _chain_eval()
+
     return app._crew_result
 
 

lib/cli/tests/test_crew_run_tui.py

@@ -137,6 +137,37 @@ def test_chain_deploy_does_not_login_for_deploy_exit(monkeypatch, capsys) -> Non
     assert "Deploy failed with exit code 42" in capsys.readouterr().out
 
 
+def test_chain_eval_runs_the_command_itself(monkeypatch, capsys) -> None:
+    """The button must not reimplement `crewai eval`, or the two can disagree."""
+    called: list[bool] = []
+    monkeypatch.setattr(
+        "crewai_cli.experimental.eval_crew.eval_crew", lambda: called.append(True)
+    )
+
+    run_crew._chain_eval()
+
+    assert called == [True]
+    assert "Evaluating this run" in capsys.readouterr().out
+
+
+def test_chain_eval_lets_the_command_say_why_it_stopped(monkeypatch, capsys) -> None:
+    """`crewai eval` exits with its own words and its own code; a clean exit is
+    not an error, and a failure is the command's to report."""
+    def refuse() -> None:
+        raise SystemExit(0)
+
+    monkeypatch.setattr("crewai_cli.experimental.eval_crew.eval_crew", refuse)
+    run_crew._chain_eval()  # no traceback, nothing added
+
+    def blow_up() -> None:
+        raise RuntimeError("AMP said no")
+
+    monkeypatch.setattr("crewai_cli.experimental.eval_crew.eval_crew", blow_up)
+    run_crew._chain_eval()  # a failed evaluation never fails the run
+
+    assert "Evaluation failed: AMP said no" in capsys.readouterr().out
+
+
 def test_view_traces_button_click_records_telemetry(monkeypatch) -> None:
     app = CrewRunApp()
     app._status = "completed"
@@ -170,6 +201,35 @@ def test_deploy_button_click_records_telemetry() -> None:
     assert exits == [app._crew_result]
 
 
+def test_evaluate_button_leaves_the_tui_and_records_telemetry() -> None:
+    """Evaluate behaves like Deploy: the app exits and the command takes over,
+    because `crewai eval` prints a link, waits, and prints a verdict."""
+    app = CrewRunApp()
+    app._status = "completed"
+    app._crew_result = object()
+    app._telemetry = Mock()
+    app._unsubscribe = lambda: None  # type: ignore[method-assign]
+    exits: list[object] = []
+    app.exit = lambda result: exits.append(result)  # type: ignore[method-assign]
+
+    app.on_button_pressed(SimpleNamespace(button=SimpleNamespace(id="btn-eval")))
+
+    app._telemetry.feature_usage_span.assert_called_once_with("cli_usage:evaluate")
+    assert app._want_eval is True
+    assert exits == [app._crew_result]
+
+
+def test_evaluate_before_completion_records_nothing() -> None:
+    app = CrewRunApp()
+    app._status = "running"
+    app._telemetry = Mock()
+
+    app.action_evaluate_crew()
+
+    app._telemetry.feature_usage_span.assert_not_called()
+    assert app._want_eval is False
+
+
 def test_conversation_turn_done_records_assistant_message() -> None:
     class RawResult:
         raw = "hello from the flow"

lib/cli/tests/test_run_declarative_flow.py

@@ -418,7 +418,7 @@ def test_id_restore_still_drops_unknown_keys(
 # ── TUI vs terminal (headless/deploy) routing ──────────────────────
 
 
-def _install_fake_flow_app(monkeypatch, *, status, want_deploy=False):
+def _install_fake_flow_app(monkeypatch, *, status, want_deploy=False, want_eval=False):
     """Replace CrewRunApp/EventListener/summary so _run_declarative_flow_tui is
     driven by a controllable fake app."""
 
@@ -430,6 +430,7 @@ def _install_fake_flow_app(monkeypatch, *, status, want_deploy=False):
             self._crew_name = crew_name
             self._status = status
             self._want_deploy = want_deploy
+            self._want_eval = want_eval
             self._crew_result = "result"
 
         def run(self):
@@ -541,6 +542,19 @@ def test_run_declarative_flow_tui_no_deploy_when_not_requested(
     assert deploy_calls == []
 
 
+def test_run_declarative_flow_tui_chains_eval(monkeypatch: pytest.MonkeyPatch) -> None:
+    """A declarative flow has the button too, and the same tail runs it."""
+    _install_fake_flow_app(monkeypatch, status="completed", want_eval=True)
+    eval_calls: list[bool] = []
+    monkeypatch.setattr("crewai_cli.run_crew._chain_eval", lambda: eval_calls.append(True))
+
+    run_declarative_flow_module._run_declarative_flow_tui(
+        SimpleNamespace(name="Flow"), None
+    )
+
+    assert eval_calls == [True]
+
+
 def test_run_declarative_flow_tui_enables_flow_events(
     monkeypatch: pytest.MonkeyPatch,
 ) -> None:

uv.lock

@@ -17,7 +17,7 @@ resolution-markers = [
 ]
 
 [options]
-exclude-newer = "2026-09-15T05:55:17.52834Z"
+exclude-newer = "0001-01-01T00:00:00Z" # This has no effect and is included for backwards compatibility when using relative exclude-newer values.
 exclude-newer-span = "P3D"
 
 [options.exclude-newer-package]
@@ -1552,7 +1552,7 @@ requires-dist = [
     { name = "httpx-auth", marker = "extra == 'a2a'", specifier = "~=0.23.1" },
     { name = "httpx-sse", marker = "extra == 'a2a'", specifier = "~=0.4.0" },
     { name = "ibm-watsonx-ai", marker = "extra == 'watson'", specifier = "~=1.3.39" },
-    { name = "instructor", specifier = ">=1.3.3" },
+    { name = "instructor", specifier = ">=1.3.3,<1.16" },
     { name = "json-repair", specifier = "~=0.60.1" },
     { name = "json5", specifier = "~=0.10.0" },
     { name = "jsonref", specifier = "~=1.1.0" },