diff --git a/src/tether/runtime/server.py b/src/tether/runtime/server.py
index fabe005..119b402 100644
--- a/src/tether/runtime/server.py
+++ b/src/tether/runtime/server.py
@@ -19,6 +19,7 @@
 from __future__ import annotations
 
 import base64
+import contextvars
 import hashlib
 import inspect
 import io
@@ -43,6 +44,7 @@
     compute_config_hash,
     compute_model_hash,
 )
+from .auth import generate_request_id
 from .tracing import get_tracer, setup_tracing, shutdown_tracing
 
 # Optional Prometheus metrics — gated on the [serve] extra (prometheus-client).
@@ -78,6 +80,12 @@ def track_in_flight(*args, **kwargs):
 
 logger = logging.getLogger(__name__)
 _tracer = get_tracer(__name__)
+REQUEST_ID_HEADER = "X-Tether-Request-ID"
+REQUEST_ID_ALIASES = (REQUEST_ID_HEADER, "X-Request-ID")
+_request_id_var: contextvars.ContextVar[str] = contextvars.ContextVar(
+    "tether_request_id",
+    default="",
+)
 
 try:
     from tether import __version__ as _TETHER_VERSION
@@ -97,6 +105,16 @@ def _coerce_optional_float(value: Any) -> float | None:
     return out
 
 
+def _resolve_http_request_id(request: Any) -> str:
+    for header in REQUEST_ID_ALIASES:
+        value = request.headers.get(header)
+        if value:
+            value = value.strip()
+            if value:
+                return value[:128]
+    return generate_request_id()
+
+
 def _call_accepts_keyword(fn: Any, keyword: str) -> bool:
     try:
         params = inspect.signature(fn).parameters
@@ -3014,6 +3032,18 @@ async def _heartbeat_loop():
         lifespan=lifespan,
     )
 
+    @app.middleware("http")
+    async def _request_id_middleware(request, call_next):
+        req_id = _resolve_http_request_id(request)
+        request.state.request_id = req_id
+        token = _request_id_var.set(req_id)
+        try:
+            response = await call_next(request)
+        finally:
+            _request_id_var.reset(token)
+        response.headers[REQUEST_ID_HEADER] = req_id
+        return response
+
     # Bearer auth dependency (Phase 1 auth-bearer feature).
     # If api_key is set at app-creation time, every protected route requires
     # the caller to pass `Authorization: Bearer <token>` (preferred) OR the
@@ -3108,6 +3138,9 @@ async def act(request: PredictRequest, _auth: None = Depends(_require_api_key)):
             # Non-standard attrs under gen_ai.action.* — proposed for upstream
             # OTel GenAI working group contribution (Phase 2 per spec).
             span.set_attribute("gen_ai.action.embodiment", _emb_label)
+            _req_id = _request_id_var.get()
+            if _req_id:
+                span.set_attribute("tether.request_id", _req_id)
             # chunk_size + denoise_steps are set AFTER predict returns (we don't
             # know them until the result is in hand). See ~line 1590 below.
             span.set_attribute(
diff --git a/tests/test_request_id_header.py b/tests/test_request_id_header.py
new file mode 100644
index 0000000..6d84b8f
--- /dev/null
+++ b/tests/test_request_id_header.py
@@ -0,0 +1,100 @@
+"""Request-id middleware coverage for the real Tether FastAPI app."""
+from __future__ import annotations
+
+from pathlib import Path
+
+import pytest
+from fastapi.testclient import TestClient
+
+
+class _StubServer:
+    def __init__(self, export_dir, *args, **kwargs):
+        self.export_dir = Path(export_dir)
+        self._ready = True
+        self.health_state = "ready"
+        self._inference_mode = "stub"
+        self._vlm_loaded = False
+        self.consecutive_crash_count = 0
+        self.max_consecutive_crashes = 5
+        self.robot_id = ""
+
+    @property
+    def ready(self):
+        return self._ready
+
+    async def load(self):
+        self._ready = True
+        self.health_state = "ready"
+
+
+@pytest.fixture
+def app(tmp_path, monkeypatch):
+    from tether.runtime import server as runtime_server
+
+    monkeypatch.setattr(runtime_server, "TetherServer", _StubServer)
+    export_dir = tmp_path / "export"
+    export_dir.mkdir()
+    app = runtime_server.create_app(str(export_dir), device="cpu")
+
+    @app.get("/_test/request-id")
+    async def request_id_probe():
+        return {"request_id": runtime_server._request_id_var.get()}
+
+    return app
+
+
+@pytest.fixture
+def client(app):
+    return TestClient(app)
+
+
+def test_health_response_has_tether_request_id_header(client):
+    response = client.get("/health")
+
+    assert response.headers["X-Tether-Request-ID"].startswith("req-")
+    assert "X-Reflex-Request-ID" not in response.headers
+
+
+def test_each_request_gets_a_unique_generated_id(client):
+    ids = {client.get("/health").headers["X-Tether-Request-ID"] for _ in range(5)}
+
+    assert len(ids) == 5
+
+
+def test_tether_request_id_header_is_echoed(client):
+    response = client.get(
+        "/health",
+        headers={"X-Tether-Request-ID": "  req-user-supplied  "},
+    )
+
+    assert response.headers["X-Tether-Request-ID"] == "req-user-supplied"
+
+
+def test_generic_request_id_header_is_accepted_when_tether_header_missing(client):
+    response = client.get("/health", headers={"X-Request-ID": "edge-proxy-123"})
+
+    assert response.headers["X-Tether-Request-ID"] == "edge-proxy-123"
+
+
+def test_tether_header_wins_over_generic_request_id(client):
+    response = client.get(
+        "/health",
+        headers={
+            "X-Tether-Request-ID": "req-tether",
+            "X-Request-ID": "proxy-request",
+        },
+    )
+
+    assert response.headers["X-Tether-Request-ID"] == "req-tether"
+
+
+def test_request_id_is_available_inside_route_context(app):
+    client = TestClient(app)
+
+    response = client.get(
+        "/_test/request-id",
+        headers={"X-Tether-Request-ID": "req-route-context"},
+    )
+
+    assert response.json()["request_id"] == "req-route-context"
+    assert response.headers["X-Tether-Request-ID"] == "req-route-context"