fix(cli): skip wrapper-side MCP discovery when chat launches the TUI

Each TUI instance spawned three stdio MCP server copies: one in the
CLI wrapper, one in tui_gateway.entry, one in the slash worker. The
wrapper's copy is dead weight — _launch_tui blocks in subprocess.call
until the TUI exits, so its registered MCP tools are never invoked,
yet the server process (35-85 MB) lives for the whole session.

Root cause: _is_tui_chat_launch() only detected --tui / HERMES_TUI=1,
so bare `hermes` with display.interface: tui fell through to
background MCP discovery in the wrapper while the TUI gateway
(spawned moments later) ran a second discovery.

Fix: _is_tui_chat_launch() now consults _resolve_use_tui() — the exact
TUI-vs-classic decision cmd_chat makes — for chat commands only
(command in {None, "chat"}), leaving mcp serve / gateway / acp / cron
discovery behavior untouched.

Verified: unit tests (RED->GREEN); E2E with a canary stdio MCP server
in a scratch HERMES_HOME counted 2 spawned copies pre-fix vs 1
post-fix (gateway's only), and the wrapper's RSS dropped ~43 MB.

Related: #71928 (same per-process duplication class), #11115 (lazy
non-core discovery).

hermes_cli/main.py

@@ -12791,7 +12791,19 @@ _AGENT_SUBCOMMANDS = {
 
 
 def _is_tui_chat_launch(args) -> bool:
-    return bool(getattr(args, "tui", False) or os.environ.get("HERMES_TUI") == "1")
+    if getattr(args, "tui", False) or os.environ.get("HERMES_TUI") == "1":
+        return True
+    # The chat path decides TUI-vs-classic via _resolve_use_tui (--cli/--tui
+    # flags, TTY gate, HERMES_TUI env, display.interface config). Bare
+    # `hermes`/`hermes chat` with a TUI display config was previously missed
+    # here, so the wrapper pre-warmed its own MCP discovery while the TUI
+    # gateway (spawned moments later) ran a second one — an idle stdio MCP
+    # server copy held dead for the whole session. Only chat commands can
+    # launch the TUI; other commands (mcp serve, gateway, acp, cron) keep
+    # their own discovery behavior untouched.
+    if getattr(args, "command", None) not in {None, "chat"}:
+        return False
+    return _resolve_use_tui(args)
 
 
 def _command_has_dedicated_mcp_startup(args) -> bool:

tests/hermes_cli/test_mcp_startup.py

@@ -102,6 +102,96 @@ def test_prepare_agent_startup_backgrounds_blocking_mcp_for_chat(monkeypatch):
         stop.set()
 
 
+def test_prepare_agent_startup_skips_discovery_when_chat_resolves_to_tui(
+    monkeypatch,
+):
+    """Bare ``hermes`` / ``hermes chat`` on a TTY with ``display.interface:
+    tui`` resolves to the TUI via ``_resolve_use_tui``, but does NOT pass
+    ``--tui`` or ``HERMES_TUI``. Discovery must be skipped in the wrapper:
+    the TUI gateway owns it, and the wrapper would otherwise hold a dead
+    MCP server for the entire session (3 copies per TUI instance).
+    """
+    calls = {"background": 0, "inline": 0}
+
+    monkeypatch.setattr(main_mod, "_resolve_use_tui", lambda _args: True)
+    monkeypatch.setattr(
+        mcp_startup,
+        "start_background_mcp_discovery",
+        lambda **_kwargs: calls.__setitem__("background", calls["background"] + 1),
+    )
+    monkeypatch.setitem(
+        sys.modules,
+        "hermes_cli.plugins",
+        types.SimpleNamespace(discover_plugins=lambda: None),
+    )
+    monkeypatch.setitem(
+        sys.modules,
+        "hermes_cli.config",
+        types.SimpleNamespace(
+            read_raw_config=lambda: {"mcp_servers": {"demo": {"transport": "stdio"}}},
+            load_config=lambda: {},
+        ),
+    )
+    monkeypatch.setitem(
+        sys.modules,
+        "agent.shell_hooks",
+        types.SimpleNamespace(register_from_config=lambda *_a, **_k: None),
+    )
+    monkeypatch.setitem(
+        sys.modules,
+        "tools.mcp_tool",
+        types.SimpleNamespace(
+            discover_mcp_tools=lambda: calls.__setitem__("inline", calls["inline"] + 1),
+        ),
+    )
+
+    main_mod._prepare_agent_startup(_agent_args(command=None))
+
+    assert calls["background"] == 0
+    assert calls["inline"] == 0
+    assert mcp_startup._mcp_discovery_thread is None
+
+
+def test_prepare_agent_startup_keeps_discovery_for_non_chat_commands(
+    monkeypatch,
+):
+    """Non-chat commands never launch the TUI, so they must keep their own
+    MCP discovery even when the ambient display config resolves to TUI —
+    ``_is_tui_chat_launch`` must not consult ``_resolve_use_tui`` there."""
+    calls = {"inline": 0}
+
+    monkeypatch.setattr(main_mod, "_resolve_use_tui", lambda _args: True)
+    monkeypatch.setitem(
+        sys.modules,
+        "hermes_cli.plugins",
+        types.SimpleNamespace(discover_plugins=lambda: None),
+    )
+    monkeypatch.setitem(
+        sys.modules,
+        "hermes_cli.config",
+        types.SimpleNamespace(
+            read_raw_config=lambda: {"mcp_servers": {"demo": {"transport": "stdio"}}},
+            load_config=lambda: {},
+        ),
+    )
+    monkeypatch.setitem(
+        sys.modules,
+        "agent.shell_hooks",
+        types.SimpleNamespace(register_from_config=lambda *_a, **_k: None),
+    )
+    monkeypatch.setitem(
+        sys.modules,
+        "tools.mcp_tool",
+        types.SimpleNamespace(
+            discover_mcp_tools=lambda: calls.__setitem__("inline", calls["inline"] + 1),
+        ),
+    )
+
+    main_mod._prepare_agent_startup(_agent_args(command="mcp", mcp_action="serve"))
+
+    assert calls["inline"] == 1
+
+
 def test_background_mcp_discovery_suppresses_interactive_oauth(monkeypatch):
     state = {"active": False, "during_discover": None}