Apply the request body limit to the SSE and OAuth endpoints

docs/migration.md

@@ -755,7 +755,7 @@ Transport-specific parameters have been moved off the `MCPServer` constructor an
 - `sse_path`, `message_path` - SSE transport paths, on `run(transport="sse", ...)` and `sse_app()`
 - `streamable_http_path` - StreamableHTTP endpoint path, on `run(transport="streamable-http", ...)` and `streamable_http_app()`
 - `json_response`, `stateless_http` - StreamableHTTP behavior, same two places; each also removes a server-to-client channel, see [Server-initiated sampling, elicitation, and roots raise `NoBackChannelError`](#server-initiated-sampling-elicitation-and-roots-raise-nobackchannelerror)
-- `max_request_body_size` - StreamableHTTP request-body limit, same two places
+- `max_request_body_size` - HTTP request-body limit, on `run()` for both HTTP transports and on both app methods
 - `event_store`, `retry_interval` - StreamableHTTP event handling, same two places
 - `transport_security` - DNS rebinding protection, on `run()` for both HTTP transports and on both app methods
 

docs/run/index.md

@@ -67,7 +67,7 @@ Each transport has its own keyword arguments, all on `run()`:
 * `streamable_http_path`: where the MCP endpoint lives. Default `/mcp`.
 * `json_response=True`: answer each POST with a single JSON body instead of an SSE stream. That body has room for the response and nothing else, so a tool that calls back into the client mid-request (`ctx.elicit()`, sampling) raises `NoBackChannelError` on this leg, and notifications tied to the in-flight call (progress from `ctx.report_progress()`, per-call log messages) are dropped; the standalone `GET` stream still carries unrelated ones.
 * `stateless_http=True`: a fresh transport per request, no session tracking.
-* `max_request_body_size`: largest accepted POST body in bytes. Defaults to 4 MiB; larger requests
+* `max_request_body_size`: largest accepted request body in bytes. Defaults to 4 MiB; larger requests
   receive HTTP 413 before parsing or session creation. Raise it only when legitimate MCP messages
   exceed that size.
 * `event_store`, `retry_interval`, `transport_security`: resumability and DNS-rebinding protection. They can wait, until you deploy somewhere other than localhost; **[Deploy & scale](deploy.md)** covers `transport_security`.

src/mcp/server/auth/routes.py

@@ -17,6 +17,7 @@
 from mcp.server.auth.middleware.client_auth import ClientAuthenticator
 from mcp.server.auth.provider import OAuthAuthorizationServerProvider
 from mcp.server.auth.settings import ClientRegistrationOptions, RevocationOptions
+from mcp.server.transport_security import DEFAULT_MAX_REQUEST_BODY_SIZE, RequestBodyLimitMiddleware
 from mcp.shared.auth import JWT_BEARER_GRANT_TYPE, OAuthMetadata, ProtectedResourceMetadata
 from mcp.shared.inbound import MCP_PROTOCOL_VERSION_HEADER
 
@@ -51,17 +52,24 @@ def validate_issuer_url(url: AnyHttpUrl):
 ID_JAG_GRANT_PROFILE = "urn:ietf:params:oauth:grant-profile:id-jag"
 
 
-def cors_middleware(
-    handler: Callable[[Request], Response | Awaitable[Response]],
-    allow_methods: list[str],
-) -> ASGIApp:
-    cors_app = CORSMiddleware(
-        app=request_response(handler),
+def _cors(app: ASGIApp, allow_methods: list[str]) -> ASGIApp:
+    return CORSMiddleware(
+        app=app,
         allow_origins="*",
         allow_methods=allow_methods,
         allow_headers=[MCP_PROTOCOL_VERSION_HEADER],
     )
-    return cors_app
+
+
+def _body_limited(app: ASGIApp) -> ASGIApp:
+    return RequestBodyLimitMiddleware(app, DEFAULT_MAX_REQUEST_BODY_SIZE)
+
+
+def cors_middleware(
+    handler: Callable[[Request], Response | Awaitable[Response]],
+    allow_methods: list[str],
+) -> ASGIApp:
+    return _cors(request_response(handler), allow_methods)
 
 
 def create_auth_routes(
@@ -84,11 +92,13 @@ def create_auth_routes(
         supports_identity_assertion=identity_assertion_enabled,
     )
     client_authenticator = ClientAuthenticator(provider)
+    token_handler = TokenHandler(provider, client_authenticator, identity_assertion_enabled=identity_assertion_enabled)
 
     # Create routes
     # Allow CORS requests for endpoints meant to be hit by the OAuth client
     # (with the client secret). This is intended to support things like MCP Inspector,
-    # where the client runs in a web browser.
+    # where the client runs in a web browser. CORS is the outermost wrapper so that
+    # responses produced by inner layers (such as a 413) still carry CORS headers.
     routes = [
         Route(
             "/.well-known/oauth-authorization-server",
@@ -102,17 +112,12 @@ def create_auth_routes(
             AUTHORIZATION_PATH,
             # do not allow CORS for authorization endpoint;
             # clients should just redirect to this
-            endpoint=AuthorizationHandler(provider).handle,
+            endpoint=_body_limited(request_response(AuthorizationHandler(provider).handle)),
             methods=["GET", "POST"],
         ),
         Route(
             TOKEN_PATH,
-            endpoint=cors_middleware(
-                TokenHandler(
-                    provider, client_authenticator, identity_assertion_enabled=identity_assertion_enabled
-                ).handle,
-                ["POST", "OPTIONS"],
-            ),
+            endpoint=_cors(_body_limited(request_response(token_handler.handle)), ["POST", "OPTIONS"]),
             methods=["POST", "OPTIONS"],
         ),
     ]
@@ -125,10 +130,7 @@ def create_auth_routes(
         routes.append(
             Route(
                 REGISTRATION_PATH,
-                endpoint=cors_middleware(
-                    registration_handler.handle,
-                    ["POST", "OPTIONS"],
-                ),
+                endpoint=_cors(_body_limited(request_response(registration_handler.handle)), ["POST", "OPTIONS"]),
                 methods=["POST", "OPTIONS"],
             )
         )
@@ -138,10 +140,7 @@ def create_auth_routes(
         routes.append(
             Route(
                 REVOCATION_PATH,
-                endpoint=cors_middleware(
-                    revocation_handler.handle,
-                    ["POST", "OPTIONS"],
-                ),
+                endpoint=_cors(_body_limited(request_response(revocation_handler.handle)), ["POST", "OPTIONS"]),
                 methods=["POST", "OPTIONS"],
             )
         )

src/mcp/server/lowlevel/server.py

@@ -65,12 +65,8 @@ async def main():
 from mcp.server.models import InitializationOptions
 from mcp.server.runner import serve_dual_era_loop
 from mcp.server.streamable_http import EventStore
-from mcp.server.streamable_http_manager import (
-    DEFAULT_MAX_REQUEST_BODY_SIZE,
-    StreamableHTTPASGIApp,
-    StreamableHTTPSessionManager,
-)
-from mcp.server.transport_security import TransportSecuritySettings
+from mcp.server.streamable_http_manager import StreamableHTTPASGIApp, StreamableHTTPSessionManager
+from mcp.server.transport_security import DEFAULT_MAX_REQUEST_BODY_SIZE, TransportSecuritySettings
 from mcp.shared._stream_protocols import ReadStream, WriteStream
 from mcp.shared.exceptions import MCPDeprecationWarning
 from mcp.shared.message import SessionMessage

src/mcp/server/mcpserver/server.py

@@ -87,9 +87,9 @@
 from mcp.server.sse import SseServerTransport
 from mcp.server.stdio import stdio_server
 from mcp.server.streamable_http import EventStore
-from mcp.server.streamable_http_manager import DEFAULT_MAX_REQUEST_BODY_SIZE, StreamableHTTPSessionManager
+from mcp.server.streamable_http_manager import StreamableHTTPSessionManager
 from mcp.server.subscriptions import InMemorySubscriptionBus, ListenHandler, SubscriptionBus
-from mcp.server.transport_security import TransportSecuritySettings
+from mcp.server.transport_security import DEFAULT_MAX_REQUEST_BODY_SIZE, TransportSecuritySettings
 from mcp.shared.exceptions import MCPError
 from mcp.shared.uri_template import UriTemplate
 
@@ -365,6 +365,7 @@ def run(
         port: int = ...,
         sse_path: str = ...,
         message_path: str = ...,
+        max_request_body_size: int = ...,
         transport_security: TransportSecuritySettings | None = ...,
     ) -> None: ...
 
@@ -1031,6 +1032,7 @@ async def run_sse_async(  # pragma: no cover
         port: int = 8000,
         sse_path: str = "/sse",
         message_path: str = "/messages/",
+        max_request_body_size: int = DEFAULT_MAX_REQUEST_BODY_SIZE,
         transport_security: TransportSecuritySettings | None = None,
     ) -> None:
         """Run the server using SSE transport."""
@@ -1039,6 +1041,7 @@ async def run_sse_async(  # pragma: no cover
         starlette_app = self.sse_app(
             sse_path=sse_path,
             message_path=message_path,
+            max_request_body_size=max_request_body_size,
             transport_security=transport_security,
             host=host,
         )
@@ -1093,6 +1096,7 @@ def sse_app(
         *,
         sse_path: str = "/sse",
         message_path: str = "/messages/",
+        max_request_body_size: int = DEFAULT_MAX_REQUEST_BODY_SIZE,
         transport_security: TransportSecuritySettings | None = None,
         host: str = "127.0.0.1",
     ) -> Starlette:
@@ -1105,7 +1109,9 @@ def sse_app(
                 allowed_origins=["http://127.0.0.1:*", "http://localhost:*", "http://[::1]:*"],
             )
 
-        sse = SseServerTransport(message_path, security_settings=transport_security)
+        sse = SseServerTransport(
+            message_path, security_settings=transport_security, max_request_body_size=max_request_body_size
+        )
 
         async def handle_sse(scope: Scope, receive: Receive, send: Send):  # pragma: no cover
             # Add client ID from auth context into request context if available

src/mcp/server/sse.py

@@ -52,6 +52,8 @@ async def handle_sse(request):
 
 from mcp.server.auth.middleware.bearer_auth import AuthenticatedUser, AuthorizationContext, authorization_context
 from mcp.server.transport_security import (
+    DEFAULT_MAX_REQUEST_BODY_SIZE,
+    RequestBodyLimitMiddleware,
     TransportSecurityMiddleware,
     TransportSecuritySettings,
 )
@@ -79,14 +81,22 @@ class SseServerTransport:
     _session_owners: dict[UUID, AuthorizationContext]
     _security: TransportSecurityMiddleware
 
-    def __init__(self, endpoint: str, security_settings: TransportSecuritySettings | None = None) -> None:
+    def __init__(
+        self,
+        endpoint: str,
+        security_settings: TransportSecuritySettings | None = None,
+        max_request_body_size: int = DEFAULT_MAX_REQUEST_BODY_SIZE,
+    ) -> None:
         """Creates a new SSE server transport, which will direct the client to POST
         messages to the relative path given.
 
         Args:
             endpoint: A relative path where messages should be posted
                     (e.g., "/messages/").
             security_settings: Optional security settings for DNS rebinding protection.
+            max_request_body_size: Maximum size in bytes for POSTed message bodies. Requests that
+                declare or stream a larger body receive HTTP 413. Defaults to 4 MiB, matching
+                `StreamableHTTPSessionManager`.
 
         Note:
             We use relative paths instead of full URLs for several reasons:
@@ -103,6 +113,9 @@ def __init__(self, endpoint: str, security_settings: TransportSecuritySettings |
 
         super().__init__()
 
+        if max_request_body_size <= 0:
+            raise ValueError("max_request_body_size must be a positive number of bytes")
+
         # Validate that endpoint is a relative path and not a full URL
         if "://" in endpoint or endpoint.startswith("//") or "?" in endpoint or "#" in endpoint:
             raise ValueError(
@@ -118,6 +131,7 @@ def __init__(self, endpoint: str, security_settings: TransportSecuritySettings |
         self._read_stream_writers = {}
         self._session_owners = {}
         self._security = TransportSecurityMiddleware(security_settings)
+        self._post_message_app = RequestBodyLimitMiddleware(self._handle_post_message, max_request_body_size)
         logger.debug(f"SseServerTransport initialized with endpoint: {endpoint}")
 
     @asynccontextmanager
@@ -203,6 +217,17 @@ async def response_wrapper(scope: Scope, receive: Receive, send: Send):
             self._session_owners.pop(session_id, None)
 
     async def handle_post_message(self, scope: Scope, receive: Receive, send: Send) -> None:
+        """ASGI application for the message endpoint.
+
+        Only POST is accepted (other methods get 405), and bodies larger than
+        `max_request_body_size` are answered with 413 before the message is handled.
+        """
+        if scope["method"] != "POST":
+            response = Response(status_code=405, headers={"Allow": "POST"})
+            return await response(scope, receive, send)
+        await self._post_message_app(scope, receive, send)
+
+    async def _handle_post_message(self, scope: Scope, receive: Receive, send: Send) -> None:
         logger.debug("Handling POST message")
         request = Request(scope, receive)
 

src/mcp/server/streamable_http_manager.py

@@ -4,25 +4,25 @@
 
 import contextlib
 import logging
-from collections import deque
 from collections.abc import AsyncIterator
-from typing import TYPE_CHECKING, Any, Final
+from typing import TYPE_CHECKING, Any
 from uuid import uuid4
 
 import anyio
 from anyio.abc import TaskStatus
 from mcp_types import DEFAULT_NEGOTIATED_VERSION, INVALID_REQUEST, ErrorData, JSONRPCError
 from mcp_types.version import HANDSHAKE_PROTOCOL_VERSIONS
-from starlette.datastructures import Headers
 from starlette.requests import Request
 from starlette.responses import Response
-from starlette.types import ASGIApp, Message, Receive, Scope, Send
+from starlette.types import Receive, Scope, Send
 
 from mcp.server._streamable_http_modern import handle_modern_request
 from mcp.server.auth.middleware.bearer_auth import AuthenticatedUser, AuthorizationContext, authorization_context
 from mcp.server.connection import Connection
 from mcp.server.runner import serve_connection, serve_loop
 from mcp.server.streamable_http import MCP_SESSION_ID_HEADER, EventStore, StreamableHTTPServerTransport
+from mcp.server.transport_security import DEFAULT_MAX_REQUEST_BODY_SIZE as DEFAULT_MAX_REQUEST_BODY_SIZE
+from mcp.server.transport_security import RequestBodyLimitMiddleware as RequestBodyLimitMiddleware
 from mcp.server.transport_security import TransportSecuritySettings
 from mcp.shared._compat import resync_tracer
 from mcp.shared.inbound import MCP_PROTOCOL_VERSION_HEADER
@@ -34,9 +34,6 @@
 
 logger = logging.getLogger(__name__)
 
-DEFAULT_MAX_REQUEST_BODY_SIZE: Final = 4 * 1024 * 1024
-"""Default maximum Streamable HTTP request body size in bytes (4 MiB)."""
-
 
 class StreamableHTTPSessionManager:
     """Manages StreamableHTTP sessions with optional resumability via event store.
@@ -70,7 +67,7 @@ class StreamableHTTPSessionManager:
             retry_interval is also configured, ensure the idle timeout comfortably exceeds the retry interval to
             avoid reaping sessions during normal SSE polling gaps. Default is None (no timeout). A value of 1800
             (30 minutes) is recommended for most deployments.
-        max_request_body_size: Maximum size in bytes for Streamable HTTP POST request bodies. Requests that
+        max_request_body_size: Maximum size in bytes for Streamable HTTP request bodies. Requests that
             exceed this limit receive a 413 response before parsing or session creation. Defaults to 4 MiB.
     """
 
@@ -371,66 +368,6 @@ async def run_server(*, task_status: TaskStatus[None] = anyio.TASK_STATUS_IGNORE
             await response(scope, receive, send)
 
 
-class RequestBodyLimitMiddleware:
-    """Reject oversized HTTP request bodies before invoking an ASGI application."""
-
-    def __init__(self, app: ASGIApp, max_body_size: int) -> None:
-        self.app = app
-        self.max_body_size = max_body_size
-
-    async def __call__(self, scope: Scope, receive: Receive, send: Send) -> None:
-        if scope["type"] != "http" or scope["method"] != "POST":
-            await self.app(scope, receive, send)
-            return
-
-        headers = Headers(scope=scope)
-        content_length = headers.get("content-length")
-        if content_length is not None:
-            try:
-                declared_size = int(content_length)
-            except ValueError:
-                pass
-            else:
-                if declared_size > self.max_body_size:
-                    response = Response("Request body too large", status_code=413)
-                    return await response(scope, receive, send)
-
-        received_body = bytearray()
-        received_request = False
-        body_complete = False
-        trailing_message: Message | None = None
-        while True:
-            message = await receive()
-            if message["type"] != "http.request":
-                trailing_message = message
-                break
-
-            received_request = True
-            body = message.get("body", b"")
-            if len(received_body) + len(body) > self.max_body_size:
-                response = Response("Request body too large", status_code=413)
-                return await response(scope, receive, send)
-            received_body.extend(body)
-            if not message.get("more_body", False):
-                body_complete = True
-                break
-
-        cached_messages: deque[Message] = deque()
-        if received_request:
-            cached_messages.append(
-                {"type": "http.request", "body": bytes(received_body), "more_body": not body_complete}
-            )
-        if trailing_message is not None:
-            cached_messages.append(trailing_message)
-
-        async def replay() -> Message:
-            if cached_messages:
-                return cached_messages.popleft()
-            return await receive()
-
-        await self.app(scope, replay, send)
-
-
 class StreamableHTTPASGIApp:
     """ASGI application for Streamable HTTP server transport."""