feat(apps): render running MCP Apps end-to-end (#1370) (#1403)

* feat(apps): render running MCP Apps end-to-end (#1370)

Wire the v2 Apps screen so clicking Open App actually runs an MCP App.
Previously App.tsx stubbed the three load-bearing pieces (sandbox URL,
bridge factory, action handlers), so the iframe loaded about:blank and
nothing crossed the bridge.

- Add useSandboxUrl hook (core/react) to read sandboxUrl from GET
  /api/config and thread it down as sandboxPath; render an "MCP Apps
  unavailable" state instead of a blank iframe when absent.
- Add createAppBridgeFactory: builds a host-side ext-apps AppBridge over
  the active client's SDK client, reads the tool's UI resource on the
  sandbox handshake and pushes it into the proxy, handles openLinks, and
  connects a PostMessageTransport.
- AppRenderer: pass the tool to the bridge factory; buffer tool
  input/result and release them once the view signals `initialized`
  (input before result, latest-wins).
- Move the AppRenderer handle ref to App.tsx so onOpenApp issues
  tools/call with the form values and delivers input + result to the
  running app via the bridge; onCloseApp tears the bridge/iframe down.
- Stories: themed sandbox placeholder (visible in light/dark), an
  echo-bridge running story, and an unavailable story. New unit tests
  for the factory, the buffering, useSandboxUrl, and the empty state.

The listChanged indicator badge wiring is tracked separately in #1402.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(apps): make MCP Apps render in dev — single bridge + result passthrough

Two bugs kept apps (e.g. NightRider's MIDI Pattern Editor) stuck on an empty
shell even though the host fetched the resource and called the tool correctly.

1. StrictMode double-mounted the bridge. The dev-only effect setup→cleanup→
   setup built TWO host bridges/transports, double-loaded the sandbox, and
   raced the app's ui/initialize handshake (it timed out, so `initialized`
   never fired and buffered input/result were never flushed). AppRenderer now
   defers bridge disposal to a microtask and, when the effect re-runs
   synchronously with identical inputs, cancels the disposal and reuses the
   live bridge — so a single bridge survives the double-invoke. The pending
   input/result buffer is no longer cleared on a transient cleanup.

2. The SDK rejected the server's result. callTool() validates structuredContent
   against the tool's outputSchema and throws on a mismatch — but an MCP App is
   a passthrough: the host must forward whatever the server returned to the
   running view (which legacy hosts render fine). InspectorClient.callTool gains
   a `skipOutputValidation` option that routes through client.request() (no
   wrapper validation); onOpenApp uses it.

Verified end-to-end against a real app via Playwright: the editor now renders
its grid + pattern data identically to the v1 inspector.

Tests: AppRenderer StrictMode single-bridge/no-dispose case; a get_temp_extra
test-server fixture (returns an extra structuredContent property) with
integration tests covering both the throwing default path and the
skipOutputValidation passthrough.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* feat(apps): warn (toast) when an app's result violates its outputSchema

The skipOutputValidation passthrough keeps schema-mismatched apps rendering,
but that silently hid a real spec problem from server developers. Now the
host still renders, but surfaces the mismatch.

- InspectorClient.callTool runs the SDK's own Ajv output validation on the
  skipOutputValidation path NON-fatally and returns the message as
  `outputValidationError` on the ToolCallInvocation (and the
  toolCallResultChange event); the result is still delivered.
- App.tsx onOpenApp shows a yellow toast when outputValidationError is set,
  explaining strict MCP clients may not render the app.

Tests: integration cases asserting outputValidationError is set for a
schema-violating result (get_temp_extra) and unset for a valid one (get_temp).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* feat(apps): summarize output-schema warning in toast + details modal

The previous toast dumped the full (very long) validation error and
overflowed. Now the toast shows a one-line summary with a "View validation
details" link that opens a modal containing the full error in a read-only,
scrollable textarea.

- New OutputValidationModal group component (+ story + tests).
- App.tsx: toast message is a summary + link; clicking it opens the modal
  populated with the tool name and the full outputValidationError.

Verified via Playwright against the NightRider MIDI editor: toast is concise,
link opens the modal with the scrollable details.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* refactor(apps): address PR review — theme stability, onError, validation polish

Responds to the automated PR review on #1403.

- Theme toggle no longer rebuilds a running app's bridge. createAppBridgeFactory
  reads the resolved theme from the DOM (Mantine's data-mantine-color-scheme) at
  bridge-build time instead of capturing isDark in the factory deps, so the
  factory identity is stable across theme flips. (#1)
- Wire AppRenderer's onError through AppsScreen -> InspectorView -> App.tsx,
  surfacing bridge/runtime failures (no client, late rejection, failed
  tools/call) as a red toast instead of a silent blank iframe. (#2)
- Extract validateToolOutput into a pure, unit-tested helper covering the
  matching / extra-property / missing-structuredContent / error / bad-schema
  cases (the no-structuredContent branch can't be produced via the SDK server,
  which rejects it before sending). (#4)
- useSandboxUrl aborts its in-flight fetch on unmount (cancelled flag) to avoid
  setState after unmount. (#5)
- Schema-mismatch toast: autoClose:false (details modal is one click away). (note)
- AppRenderer: add a state-machine comment block documenting the bridge
  lifecycle refs. (#6)
- inspectorClient: note why the `as CallToolResult` casts are safe. (note)

Tests: AppsScreen onError surface; validateToolOutput unit suite; factory theme
read from DOM. validate + test:storybook green.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* docs(apps): clarify theme attribute + useSandboxUrl fetch comments

Re-review follow-ups (non-blocking):
- createAppBridgeFactory: note that Mantine's data-mantine-color-scheme is only
  ever light/dark (auto is resolved before paint), so no `auto` branch is needed
  and matchMedia only covers an absent attribute.
- useSandboxUrl: correct the doc to say it fetches on mount AND re-fetches on
  baseUrl/authToken change (idempotent), not "once on mount".

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This commit is contained in:
Cliff Hall 2026-06-01 18:01:44 -04:00 • committed by GitHub
commit 5fb2b503f4
No known key found for this signature in database
GPG key ID: B5690EEEBB952194

View file

@ -1,5 +1,11 @@
import { useCallback, useEffect, useMemo, useRef, useState } from "react";
import { useComputedColorScheme, useMantineColorScheme } from "@mantine/core";
import {
Anchor,
Stack,
Text,
useComputedColorScheme,
useMantineColorScheme,
} from "@mantine/core";
import { notifications } from "@mantine/notifications";
import type {
InitializeResult,
@ -7,7 +13,6 @@ import type {
LoggingMessageNotification,
Tool,
} from "@modelcontextprotocol/sdk/types.js";
import type { AppBridge } from "@modelcontextprotocol/ext-apps/app-bridge";
import { InspectorClient } from "@inspector/core/mcp/index.js";
import type { JsonValue } from "@inspector/core/mcp/index.js";
import type {
@ -49,11 +54,13 @@ import { useManagedRequestorTasks } from "@inspector/core/react/useManagedReques
import { useResourceSubscriptions } from "@inspector/core/react/useResourceSubscriptions.js";
import { useMessageLog } from "@inspector/core/react/useMessageLog.js";
import { useFetchRequestLog } from "@inspector/core/react/useFetchRequestLog.js";
import { useSandboxUrl } from "@inspector/core/react/useSandboxUrl.js";
import { InspectorView } from "./components/views/InspectorView/InspectorView";
import type { ToolCallState } from "./components/screens/ToolsScreen/ToolsScreen";
import type { GetPromptState } from "./components/screens/PromptsScreen/PromptsScreen";
import type { ReadResourceState } from "./components/screens/ResourcesScreen/ResourcesScreen";
import type { BridgeFactory } from "./components/elements/AppRenderer/AppRenderer";
import type { AppRendererHandle } from "./components/elements/AppRenderer/AppRenderer";
import { createAppBridgeFactory } from "./components/elements/AppRenderer/createAppBridgeFactory";
import type { LogEntryData } from "./components/elements/LogEntry/LogEntry";
import {
ServerConfigModal,
@ -61,6 +68,7 @@ import {
} from "./components/groups/ServerConfigModal/ServerConfigModal";
import { ServerSettingsModal } from "./components/groups/ServerSettingsModal/ServerSettingsModal";
import { ConnectionInfoModal } from "./components/groups/ConnectionInfoModal/ConnectionInfoModal";
import { OutputValidationModal } from "./components/groups/OutputValidationModal/OutputValidationModal";
import type { OAuthDetails } from "./components/groups/ConnectionInfoContent/ConnectionInfoContent";
import { ServerRemoveConfirmModal } from "./components/groups/ServerRemoveConfirmModal/ServerRemoveConfirmModal";
import { buildExportFilename, downloadJsonFile } from "./lib/downloadFile";
@ -126,23 +134,6 @@ function getAuthToken(): string | undefined {
}
}
// MCP Apps sandbox — the iframe URL the parent should embed, plus the
// per-tool bridge factory. The dev backend serves `sandbox_proxy.html` on
// the sandbox controller port; the factory will eventually wrap the SDK
// client. For now neither is wired (the Apps tab uses these props but does
// not yet round-trip tool input through a live bridge — that's a follow-up
// alongside the AppRenderer integration). Keep these as stable references
// so InspectorView's effect deps don't churn.
const STUB_SANDBOX_PATH = "about:blank";
const stubBridgeFactory: BridgeFactory = () =>
({
sendToolInput: async () => {},
sendToolResult: async () => {},
sendToolCancelled: async () => {},
teardownResource: async () => ({}),
close: async () => {},
}) as unknown as AppBridge;
// Derive `LogEntryData[]` from the MessageLog by filtering for the
// `notifications/message` notifications the server emits in response to
// `logging/setLevel`. The Logs screen renders these; we transform here
@ -177,6 +168,26 @@ const EMPTY_SETTINGS: InspectorServerSettings = {
requestTimeout: 0,
};
// Body of the output-schema-mismatch warning toast: a one-line summary plus a
// link that opens the full validation details in a modal (the raw error is far
// too long for a toast).
const OutputValidationToastMessage = ({
onViewDetails,
}: {
onViewDetails: () => void;
}) => (
<Stack gap={4}>
<Text size="sm">
The tool result&apos;s structuredContent doesn&apos;t match the
tool&apos;s outputSchema. The inspector renders it anyway, but strict MCP
clients may not.
</Text>
<Anchor component="button" type="button" size="sm" onClick={onViewDetails}>
View validation details
</Anchor>
</Stack>
);
function App() {
// Theme toggle plumbing (preserved from the pre-wire placeholder).
const { setColorScheme } = useMantineColorScheme();
@ -214,6 +225,11 @@ function App() {
>(undefined);
const [connectionInfoModalOpen, setConnectionInfoModalOpen] = useState(false);
const [removeTarget, setRemoveTarget] = useState<ServerEntry | null>(null);
// Details for the output-schema-mismatch modal opened from the warning toast.
const [outputValidationDetails, setOutputValidationDetails] = useState<{
toolName: string;
message: string;
} | null>(null);
// The active connection target. `null` between sessions; set as soon as
// the user toggles a server card on. Drives state-manager lifetime.
@ -226,6 +242,33 @@ function App() {
// next switch happens (or when the component unmounts).
const [inspectorClient, setInspectorClient] =
useState<InspectorClient | null>(null);
// MCP Apps runtime wiring. `sandboxUrl` is the inspector's sandbox-proxy page
// (the trusted outer iframe); `appRendererRef` lets the app handlers push tool
// input/result into the running app and tear it down. The bridge factory wraps
// the active client's underlying SDK client so the running view can call the
// server, and reads the tool's UI resource into the sandbox on handshake.
const appRendererRef = useRef<AppRendererHandle>(null);
const { sandboxUrl } = useSandboxUrl({
baseUrl:
typeof window !== "undefined"
? window.location.origin
: "http://localhost",
authToken: getAuthToken(),
});
const sandboxBridgeFactory = useMemo(
() =>
createAppBridgeFactory({
getClient: () => inspectorClient?.getAppRendererClient() ?? null,
readResource: async (uri) => {
if (!inspectorClient) throw new Error("No MCP client connected.");
const invocation = await inspectorClient.readResource(uri);
return invocation.result;
},
}),
[inspectorClient],
);
const [managedToolsState, setManagedToolsState] =
useState<ManagedToolsState | null>(null);
const [managedPromptsState, setManagedPromptsState] =
@ -875,6 +918,88 @@ function App() {
setToolCallState(undefined);
}, []);
// --- MCP Apps handlers. Unlike onCallTool (which feeds the Tools panel),
// these route the tool input/result into the running app via the renderer's
// imperative handle. ---
// Surfaces bridge/runtime failures (factory throw — e.g. no client after a
// disconnect — late bridge rejection, or a failed tools/call) that would
// otherwise leave a silently blank app iframe.
const onAppError = useCallback((err: Error) => {
notifications.show({
title: "MCP App error",
message: err.message,
color: "red",
});
}, []);
// Selection is owned by AppsScreen's local state; App.tsx has nothing to do
// on select, but the prop is required so the screen stays prop-driven.
const onSelectApp = useCallback(() => {}, []);
const onOpenApp = useCallback(
async (name: string, args: Record<string, unknown>) => {
if (!inspectorClient) return;
const tool = tools.find((t: Tool) => t.name === name);
if (!tool) return;
// AppsScreen flips `running` -> mounts AppRenderer in the same tick it
// calls this, so the renderer handle isn't wired yet. Yield one microtask
// (after React commits the mount) before pushing input; the renderer then
// buffers it until the view's `initialized` event, releasing input before
// result.
await Promise.resolve();
void appRendererRef.current?.sendToolInput(args);
try {
// skipOutputValidation: the result is forwarded verbatim to the running
// app (the real consumer), so the host must not reject it on its own
// outputSchema validation — that would deny the app a result the server
// actually returned and legacy hosts render fine.
const invocation = await inspectorClient.callTool(
tool,
args as Record<string, JsonValue>,
undefined,
undefined,
undefined,
{ skipOutputValidation: true },
);
if (invocation.success && invocation.result) {
void appRendererRef.current?.sendToolResult(invocation.result);
}
// Leniency above keeps the app rendering, but surface the schema
// mismatch so a server developer knows strict MCP clients may refuse
// to render this app. The full validation error is too long for a
// toast, so summarize and link to a modal with the details.
if (invocation.outputValidationError) {
const details = {
toolName: tool.name,
message: invocation.outputValidationError,
};
notifications.show({
// Don't auto-dismiss: the message is advisory and the details modal
// is one click away — let the user close it when they've read it.
autoClose: false,
title: "App output doesn't match its schema",
color: "yellow",
message: (
<OutputValidationToastMessage
onViewDetails={() => setOutputValidationDetails(details)}
/>
),
});
}
} catch (err) {
// Transport-level failure (the call never returned a result). Surface it
// so the user isn't left staring at a blank/partial app frame.
onAppError(err instanceof Error ? err : new Error(String(err)));
}
},
[inspectorClient, tools, onAppError],
);
const onCloseApp = useCallback(() => {
void appRendererRef.current?.teardown();
}, []);
const onGetPrompt = useCallback(
async (name: string, args: Record<string, string>) => {
if (!inspectorClient) return;
@ -1224,8 +1349,9 @@ function App() {
getPromptState={getPromptState}
readResourceState={effectiveReadResourceState}
currentLogLevel={currentLogLevel}
sandboxPath={STUB_SANDBOX_PATH}
bridgeFactory={stubBridgeFactory}
sandboxPath={sandboxUrl}
bridgeFactory={sandboxBridgeFactory}
appRendererRef={appRendererRef}
onToggleTheme={onToggleTheme}
onToggleConnection={(id) => {
void onToggleConnection(id);
@ -1274,9 +1400,12 @@ function App() {
onTogglePinHistory={todoNoop}
onClearNetwork={onClearNetwork}
onExportNetwork={onExportNetwork}
onSelectApp={todoNoop}
onOpenApp={todoNoop}
onCloseApp={todoNoop}
onSelectApp={onSelectApp}
onOpenApp={(name, args) => {
void onOpenApp(name, args);
}}
onCloseApp={onCloseApp}
onAppError={onAppError}
onRefreshApps={onRefreshTools}
/>
<ServerConfigModal
@ -1310,6 +1439,12 @@ function App() {
onCancel={() => setRemoveTarget(null)}
onConfirm={onConfirmRemove}
/>
<OutputValidationModal
opened={outputValidationDetails !== null}
toolName={outputValidationDetails?.toolName}
message={outputValidationDetails?.message}
onClose={() => setOutputValidationDetails(null)}
/>
</>
);
}