Sieve only core-vocabulary results; claimed shapes pass through

Moving extension interception to the handler layer put interceptor
results through the per-version result sieve, which strips fields the
surface doesn't know - breaking claimed extension shapes (a resultType
outside the core vocabulary carrying vendor fields). The sieve's job is
per-version surface enforcement for core results; an extension
resultType marks a shape the extension owns, so it now applies to
core-vocabulary results only. Claimed shapes still get the 2026-era
serverInfo stamp: they are results like any other.

Author: maxisbey

src/mcp/server/runner.py

@@ -24,6 +24,7 @@
 from mcp_types import (
     CLIENT_CAPABILITIES_META_KEY,
     CLIENT_INFO_META_KEY,
+    CORE_RESULT_TYPES,
     INTERNAL_ERROR,
     INVALID_PARAMS,
     INVALID_REQUEST,
@@ -335,9 +336,10 @@ def _serialize(self, method: str, version: str, result: HandlerResult) -> dict[s
         of the inbound classification ladder.
 
         One pass owns the whole response envelope, in order: cache hints fill
-        `ttlMs`/`cacheScope` the handler left unset, spec-method results are
-        validated and sieved by the per-version surface, and 2026-era results
-        get the `serverInfo` `_meta` stamp (spec #3002). Runs inside the
+        `ttlMs`/`cacheScope` the handler left unset, core-vocabulary spec-method
+        results are validated and sieved by the per-version surface (a claimed
+        extension `resultType` shape is the extension's to own), and 2026-era
+        results get the `serverInfo` `_meta` stamp (spec #3002). Runs inside the
         middleware chain so the OpenTelemetry span observes a failing return
         shape (unsupported type, malformed spec result) as an error rather
         than closing on a request that the client sees fail - and so a
@@ -352,10 +354,15 @@ def _serialize(self, method: str, version: str, result: HandlerResult) -> dict[s
                 # Hint keys first so wire keys the handler set win, matching `apply_cache_hint` precedence.
                 result = {"ttlMs": hint.ttl_ms, "cacheScope": hint.scope, **result}
         dumped = _dump_result(result)
-        # TODO(L56): reject resultType values outside {"complete", "input_required"} unless the
-        # corresponding extension is in this request's _meta clientCapabilities.extensions; the
+        # An extension `resultType` (outside the core vocabulary) marks a claimed
+        # shape owned by the extension that defined it: the per-version surface
+        # doesn't describe it, so the sieve applies to core results only.
+        # TODO(L56): reject extension resultType values unless the corresponding
+        # extension is in this request's _meta clientCapabilities.extensions; the
         # explicit MUST-reject is client-side (basic/index.mdx ResultType), this enforces it proactively.
-        if method in _methods.SPEC_CLIENT_METHODS:
+        result_type = dumped.get("resultType")
+        core_shape = not isinstance(result_type, str) or result_type in CORE_RESULT_TYPES
+        if method in _methods.SPEC_CLIENT_METHODS and core_shape:
             try:
                 dumped = _methods.serialize_server_result(method, version, dumped)
             except ValidationError:

tests/server/test_runner.py

@@ -27,6 +27,7 @@
     PROTOCOL_VERSION_META_KEY,
     SERVER_INFO_META_KEY,
     UNSUPPORTED_PROTOCOL_VERSION,
+    CallToolRequestParams,
     ClientCapabilities,
     EmptyResult,
     ErrorData,
@@ -1059,6 +1060,27 @@ async def custom(ctx: Ctx, params: RequestParams) -> dict[str, Any]:
     assert retained == {"ok": True}
 
 
+@pytest.mark.anyio
+async def test_a_claimed_extension_result_type_bypasses_the_sieve_and_is_stamped(server: SrvT):
+    """SDK-defined: a spec-method result carrying an extension `resultType` is a
+    claimed shape the extension owns - the per-version sieve would strip its
+    vendor fields, so it applies to core-vocabulary results only. The identity
+    stamp still lands: claimed shapes are results like any other."""
+
+    async def custom(ctx: Ctx, params: CallToolRequestParams) -> dict[str, Any]:
+        return {"resultType": "voucher", "voucherCode": "v-42"}
+
+    server.add_request_handler("tools/call", CallToolRequestParams, custom)
+    born_ready = Connection.from_envelope(LATEST_MODERN_VERSION, None, None)
+    async with connected_runner(server, initialized=False, connection=born_ready) as (client, _):
+        result = await client.send_raw_request("tools/call", _modern_params(name="issue"))
+    assert result == {
+        "resultType": "voucher",
+        "voucherCode": "v-42",
+        "_meta": {SERVER_INFO_META_KEY: {"name": "test-server", "version": "0.0.1"}},
+    }
+
+
 @pytest.mark.anyio
 async def test_an_empty_result_on_the_modern_path_carries_only_the_stamp(server: SrvT):
     """Spec-mandated (2026-07-28, #3002): serverInfo is stamped into every