diff --git a/gateway/platforms/api_server.py b/gateway/platforms/api_server.py
index 796cc53a4f..0d1ec17de8 100644
--- a/gateway/platforms/api_server.py
+++ b/gateway/platforms/api_server.py
@@ -3465,7 +3465,12 @@ class APIServerAdapter(OpenAICompatRoutesMixin, BasePlatformAdapter):
             if event_type == "reasoning.available":
                 events.enqueue("tool.progress", {"message_id": message_id, "tool_name": tool_name or "_thinking", "delta": preview or ""})
             elif event_type in {"tool.started", "tool.completed", "tool.failed"}:
-                events.enqueue(event_type, {"message_id": message_id, "tool_name": tool_name, "preview": preview, "args": args})
+                event_name = (
+                    "tool.failed"
+                    if event_type == "tool.completed" and kwargs.get("is_error")
+                    else event_type
+                )
+                events.enqueue(event_name, {"message_id": message_id, "tool_name": tool_name, "preview": preview, "args": args})
 
         def _commentary(text: str, *, already_streamed: bool = False) -> None:
             # Mid-turn assistant commentary (Codex ``phase="commentary"``, text beside tool calls)
diff --git a/tests/gateway/test_session_api.py b/tests/gateway/test_session_api.py
index a3a2ae94c2..7a87cacf95 100644
--- a/tests/gateway/test_session_api.py
+++ b/tests/gateway/test_session_api.py
@@ -376,6 +376,38 @@ async def test_session_chat_stream_disconnect_keeps_control_refs_until_executor_
     assert run_id not in adapter._active_run_agents
 
 
+@pytest.mark.asyncio
+async def test_session_chat_stream_classifies_failed_tool_completions(adapter, session_db):
+    session_id = session_db.create_session("tool-status-stream", "api_server")
+
+    async def fake_run(**kwargs):
+        progress = kwargs["tool_progress_callback"]
+        progress("tool.completed", tool_name="read_file", is_error=False)
+        progress("tool.completed", tool_name="terminal", is_error=True)
+        progress("tool.failed", tool_name="web_search")
+        return {"final_response": "done", "session_id": session_id}, {"total_tokens": 1}
+
+    app = _create_session_app(adapter)
+    with patch.object(adapter, "_run_agent", side_effect=fake_run):
+        async with TestClient(TestServer(app)) as cli:
+            resp = await cli.post(
+                f"/api/sessions/{session_id}/chat/stream",
+                json={"message": "run tools"},
+            )
+            assert resp.status == 200
+            body = await resp.text()
+
+    blocks = body.split("\n\n")
+    assert any("event: tool.completed" in b and '"tool_name": "read_file"' in b for b in blocks)
+    assert any("event: tool.failed" in b and '"tool_name": "terminal"' in b for b in blocks)
+    assert any("event: tool.failed" in b and '"tool_name": "web_search"' in b for b in blocks)
+    assert body.count("event: tool.completed") == 1
+    assert body.count("event: tool.failed") == 2
+    assert '"tool_name": "read_file"' in body
+    assert '"tool_name": "terminal"' in body
+    assert '"tool_name": "web_search"' in body
+
+
 @pytest.mark.asyncio
 async def test_session_chat_stream_run_completed_carries_turn_transcript(adapter, session_db):
     """run.completed must include the full interleaved turn transcript so a
diff --git a/website/docs/user-guide/features/api-server.md b/website/docs/user-guide/features/api-server.md
index 142caf45b0..1c940e6db4 100644
--- a/website/docs/user-guide/features/api-server.md
+++ b/website/docs/user-guide/features/api-server.md
@@ -614,7 +614,7 @@ External UIs can manage Hermes sessions over REST without standing up the dashbo
 | `GET` | `/api/sessions/{id}/messages` | Message history for a session |
 | `POST` | `/api/sessions/{id}/fork` | Branch the session via `SessionDB` lineage (matches CLI `/branch` semantics) |
 | `POST` | `/api/sessions/{id}/chat` | Run one synchronous agent turn |
-| `POST` | `/api/sessions/{id}/chat/stream` | SSE wrapper over a single turn — emits `assistant.delta`, `assistant.commentary` (mid-turn commentary: `message_id`, `text`, `already_streamed`; never folded into `assistant.completed`), `tool.started`, `tool.completed`, then a terminal `run.completed` / `run.failed` / `run.cancelled` event that matches how the turn ended (see [Terminal run status](../../developer-guide/programmatic-integration.md#terminal-run-status)) |
+| `POST` | `/api/sessions/{id}/chat/stream` | SSE wrapper over a single turn — emits `assistant.delta`, `assistant.commentary` (mid-turn commentary: `message_id`, `text`, `already_streamed`; never folded into `assistant.completed`), `tool.started`, `tool.completed`, `tool.failed` (a tool that finished with an error), then a terminal `run.completed` / `run.failed` / `run.cancelled` event that matches how the turn ended (see [Terminal run status](../../developer-guide/programmatic-integration.md#terminal-run-status)) |
 
 `/v1/capabilities` advertises the full surface via `session_*` feature flags and `endpoints.session_*` entries so external UIs can detect support and fall back safely. Inline images are supported in `chat` and `chat/stream` payloads (multimodal-aware path).