witchan · pick7 · Aug 9, 2026
diff --git a/IOSMCPPreferences.h b/IOSMCPPreferences.h
@@ -6,10 +6,14 @@
 #define IOS_MCP_DEFAULT_PORT 8090
 #define IOS_MCP_MIN_PORT 1024
 #define IOS_MCP_MAX_PORT 65535
+#define IOS_MCP_DEFAULT_MAX_CONCURRENT_SCREEN_TASKS 1
+#define IOS_MCP_MIN_CONCURRENT_SCREEN_TASKS 1
+#define IOS_MCP_MAX_CONCURRENT_SCREEN_TASKS 2
 #define IOS_MCP_PREFERENCES_DOMAIN @"com.witchan.ios-mcp.preferences"
 #define IOS_MCP_ENABLED_PREFERENCE_KEY @"enabled"
 #define IOS_MCP_PORT_PREFERENCE_KEY @"port"
 #define IOS_MCP_DEBUG_LOGGING_PREFERENCE_KEY @"debugLoggingEnabled"
+#define IOS_MCP_MAX_CONCURRENT_SCREEN_TASKS_PREFERENCE_KEY @"maxConcurrentScreenTasks"
 #define IOS_MCP_DARWIN_NOTIFICATION_START CFSTR("com.witchan.ios-mcp.control/start")
 #define IOS_MCP_DARWIN_NOTIFICATION_STOP CFSTR("com.witchan.ios-mcp.control/stop")
 
@@ -50,6 +54,23 @@ static inline uint16_t IOSMCPConfiguredPort(void) {
     return port;
 }
 
+static inline NSInteger IOSMCPConfiguredMaxConcurrentScreenTasks(void) {
+    NSInteger limit = IOS_MCP_DEFAULT_MAX_CONCURRENT_SCREEN_TASKS;
+    CFPreferencesAppSynchronize((__bridge CFStringRef)IOS_MCP_PREFERENCES_DOMAIN);
+    CFPropertyListRef value = CFPreferencesCopyAppValue(
+        (__bridge CFStringRef)IOS_MCP_MAX_CONCURRENT_SCREEN_TASKS_PREFERENCE_KEY,
+        (__bridge CFStringRef)IOS_MCP_PREFERENCES_DOMAIN);
+    if (value) {
+        id preference = (__bridge id)value;
+        if ([preference respondsToSelector:@selector(integerValue)]) {
+            limit = [preference integerValue];
+        }
+        CFRelease(value);
+    }
+    return MAX(IOS_MCP_MIN_CONCURRENT_SCREEN_TASKS,
+               MIN(limit, IOS_MCP_MAX_CONCURRENT_SCREEN_TASKS));
+}
+
 static inline NSString *IOSMCPCurrentLANIPAddress(void) {
     struct ifaddrs *interfaces = NULL;
     NSString *preferredAddress = nil;

diff --git a/MCPServer.m b/MCPServer.m
@@ -10,6 +10,7 @@
 #import "LogManager.h"
 #import "OCRManager.h"
 #import "MCPLogger.h"
+#import "IOSMCPPreferences.h"
 #import <UIKit/UIKit.h>
 #import <sys/socket.h>
 #import <netinet/in.h>
@@ -655,6 +656,7 @@ - (NSDictionary *)routeMCPRequest:(NSDictionary *)request;
 - (NSDictionary *)handleInitialize:(id)reqId params:(NSDictionary *)params;
 - (NSDictionary *)handleToolsList:(id)reqId;
 - (NSDictionary *)handleToolsCall:(id)reqId params:(NSDictionary *)params;
+- (NSDictionary *)performScreenTaskForRequest:(id)reqId task:(NSDictionary *(^)(void))task;
 - (NSDictionary *)lockedScreenGuardResponseForTool:(NSString *)toolName reqId:(id)reqId;
 - (NSDictionary *)executeButtonPress:(id)reqId button:(HIDButtonType)button args:(NSDictionary *)args label:(NSString *)label;
 - (BOOL)pressButtonSynchronously:(HIDButtonType)button duration:(NSTimeInterval)duration timeout:(NSTimeInterval)timeout error:(NSString **)error;
@@ -725,6 +727,7 @@ @implementation MCPServer {
     int _serverSocket;
     dispatch_source_t _acceptSource;
     dispatch_queue_t _clientQueue;
+    dispatch_semaphore_t _screenTaskSemaphore;
     NSString *_sessionId;
     NSString *_negotiatedProtocolVersion;
 }
@@ -742,7 +745,12 @@ - (instancetype)init {
     self = [super init];
     if (self) {
         _serverSocket = -1;
-        _clientQueue = dispatch_queue_create("com.witchan.ios-mcp.client", DISPATCH_QUEUE_CONCURRENT);
+        dispatch_queue_attr_t clientAttributes =
+            dispatch_queue_attr_make_with_autorelease_frequency(DISPATCH_QUEUE_CONCURRENT,
+                                                                 DISPATCH_AUTORELEASE_FREQUENCY_WORK_ITEM);
+        _clientQueue = dispatch_queue_create("com.witchan.ios-mcp.client", clientAttributes);
+        _screenTaskSemaphore =
+            dispatch_semaphore_create(IOSMCPConfiguredMaxConcurrentScreenTasks());
         _sessionId = [[NSUUID UUID] UUIDString];
         _negotiatedProtocolVersion = MCP_PROTOCOL_VERSION_LATEST;
     }
@@ -808,7 +816,10 @@ - (void)startOnPort:(uint16_t)port {
     _port = port;
     _running = YES;
 
-    dispatch_queue_t queue = dispatch_queue_create("com.witchan.ios-mcp.accept", DISPATCH_QUEUE_CONCURRENT);
+    dispatch_queue_attr_t acceptAttributes =
+        dispatch_queue_attr_make_with_autorelease_frequency(DISPATCH_QUEUE_CONCURRENT,
+                                                             DISPATCH_AUTORELEASE_FREQUENCY_WORK_ITEM);
+    dispatch_queue_t queue = dispatch_queue_create("com.witchan.ios-mcp.accept", acceptAttributes);
     _acceptSource = dispatch_source_create(DISPATCH_SOURCE_TYPE_READ, sock, 0, queue);
 
     __weak typeof(self) weakSelf = self;
@@ -819,7 +830,9 @@ - (void)startOnPort:(uint16_t)port {
         if (client >= 0) {
             MCPSetCloseOnExec(client);
             dispatch_async(self->_clientQueue, ^{
-                [self handleClient:client];
+                @autoreleasepool {
+                    [self handleClient:client];
+                }
             });
         }
     });
@@ -2150,6 +2163,27 @@ - (NSDictionary *)handleToolsList:(id)reqId {
 
 #pragma mark - MCP: tools/call
 
+- (NSDictionary *)performScreenTaskForRequest:(id)reqId task:(NSDictionary *(^)(void))task {
+    if (!task) {
+        return [self mcpError:reqId code:-32603 message:@"Missing screen task"];
+    }
+
+    dispatch_time_t timeout = dispatch_time(DISPATCH_TIME_NOW, 30 * NSEC_PER_SEC);
+    if (dispatch_semaphore_wait(_screenTaskSemaphore, timeout) != 0) {
+        return [self mcpError:reqId code:-32000 message:@"Timed out waiting for screen processing"];
+    }
+
+    NSDictionary *result = nil;
+    @try {
+        @autoreleasepool {
+            result = task();
+        }
+    } @finally {
+        dispatch_semaphore_signal(_screenTaskSemaphore);
+    }
+    return result;
+}
+
 - (NSDictionary *)handleToolsCall:(id)reqId params:(NSDictionary *)params {
     if (![params isKindOfClass:[NSDictionary class]]) {
         return [self mcpError:reqId code:-32602 message:@"Invalid params: expected object"];
@@ -2207,7 +2241,9 @@ - (NSDictionary *)handleToolsCall:(id)reqId params:(NSDictionary *)params {
     else if ([toolName isEqualToString:@"get_screen_info"]) {
         return [self executeScreenInfo:reqId];
     } else if ([toolName isEqualToString:@"screenshot"]) {
-        return [self executeScreenshot:reqId args:args];
+        return [self performScreenTaskForRequest:reqId task:^{
+            return [self executeScreenshot:reqId args:args];
+        }];
     }
     // Clipboard tools
     else if ([toolName isEqualToString:@"get_clipboard"]) {
@@ -2233,8 +2269,17 @@ - (NSDictionary *)handleToolsCall:(id)reqId params:(NSDictionary *)params {
     } else if ([toolName isEqualToString:@"get_element_at_point"]) {
         return [self executeGetElementAtPoint:reqId args:args];
     } else if ([toolName isEqualToString:@"ocr_screen"]) {
-        return [self executeOCRScreen:reqId args:args];
+        return [self performScreenTaskForRequest:reqId task:^{
+            return [self executeOCRScreen:reqId args:args];
+        }];
     } else if ([toolName isEqualToString:@"describe_screen"]) {
+        BOOL includesImageWork = [args[@"include_ocr"] boolValue] ||
+                                 [args[@"include_screenshot"] boolValue];
+        if (includesImageWork) {
+            return [self performScreenTaskForRequest:reqId task:^{
+                return [self executeDescribeScreen:reqId args:args];
+            }];
+        }
         return [self executeDescribeScreen:reqId args:args];
     }
     // Text input tools

diff --git a/OCRManager.m b/OCRManager.m
@@ -142,9 +142,11 @@ - (NSDictionary *)recognizeTextWithLanguages:(NSArray<NSString *> *)languages
 
         // Downsample large captures before OCR (longest edge cap). Speeds up Vision on
         // high-res iPad screens; coordinates still map back via the logical image.size.
-        CGImageRef ocrImage = image.CGImage;
         CGImageRef downsampled = OCRCreateDownsampled(image.CGImage, 1600.0);
-        if (downsampled) ocrImage = downsampled;
+        CGImageRef ocrImage = downsampled ?: image.CGImage;
+        if (downsampled) {
+            image = nil;
+        }
 
         VNImageRequestHandler *handler = [[VNImageRequestHandler alloc] initWithCGImage:ocrImage options:@{}];
         NSError *performError = nil;

diff --git a/ScreenManager.m b/ScreenManager.m
@@ -14,7 +14,7 @@
 } while (0)
 
 typedef struct __IOSurface *IOSurfaceRef;
-typedef UIImage *(*UICreateScreenUIImageFunc)(void);
+typedef UIImage *(*UICreateScreenUIImageFunc)(void) NS_RETURNS_RETAINED;
 typedef CGImageRef (*UICreateCGImageFromIOSurfaceFunc)(IOSurfaceRef surface);
 typedef CGImageRef (*CARenderServerCaptureDisplayFunc)(uint32_t serverPort, CFStringRef displayName, CFDictionaryRef options);
 
@@ -27,6 +27,14 @@
 static const CGFloat kMCPScreenshotMinimumJPEGQuality = 0.30;
 static const NSInteger kMCPScreenshotJPEGSearchPasses = 6;
 
+static NSData *MCPJPEGRepresentation(UIImage *image, CGFloat quality) {
+    NSData *data = nil;
+    @autoreleasepool {
+        data = UIImageJPEGRepresentation(image, quality);
+    }
+    return data;
+}
+
 /// Point-space target size for a pixel-space capture of `pixelSize`.
 ///
 /// Screenshots are downsampled so that one image pixel equals one screen point, which makes the
@@ -295,11 +303,13 @@ - (NSDictionary *)deviceInteractionStateOnMainThread {
 - (NSDictionary *)takeScreenshotPayload {
     __block NSDictionary *payload = nil;
     dispatch_block_t block = ^{
-        payload = [self privateScreenshotPayload];
-        if (!payload) {
-            SCREEN_LOG(@"Private screenshot APIs produced no encodable image, falling back to window capture");
-            UIImage *image = [self fallbackScreenshotImage];
-            payload = [self payloadByEncodingImage:image source:@"window_capture"];
+        @autoreleasepool {
+            payload = [self privateScreenshotPayload];
+            if (!payload) {
+                SCREEN_LOG(@"Private screenshot APIs produced no encodable image, falling back to window capture");
+                UIImage *image = [self fallbackScreenshotImage];
+                payload = [self payloadByEncodingImage:image source:@"window_capture"];
+            }
         }
     };
 
@@ -315,10 +325,6 @@ - (NSDictionary *)privateScreenshotPayload {
     UIImage *image = nil;
     NSDictionary *payload = nil;
 
-    NSData *screenData = [ScreenManager getScreenDataWithQuantity:(NSInteger)round(kMCPScreenshotInitialJPEGQuality * 100.0)];
-    payload = [self payloadByEncodingImageData:screenData source:@"getScreenDataWithQuantity"];
-    if (payload) return payload;
-
     image = [self screenshotImageFromRenderServerCapture];
     payload = [self payloadByEncodingImage:image source:@"render_server"];
     if (payload) return payload;
@@ -331,6 +337,10 @@ - (NSDictionary *)privateScreenshotPayload {
     payload = [self payloadByEncodingImage:image source:@"createScreenIOSurface"];
     if (payload) return payload;
 
+    NSData *screenData = [ScreenManager getScreenDataWithQuantity:(NSInteger)round(kMCPScreenshotInitialJPEGQuality * 100.0)];
+    payload = [self payloadByEncodingImageData:screenData source:@"getScreenDataWithQuantity"];
+    if (payload) return payload;
+
     return nil;
 }
 
@@ -400,9 +410,11 @@ - (UIImage *)privateScreenshotImage {
 - (UIImage *)captureScreenImage {
     __block UIImage *image = nil;
     dispatch_block_t block = ^{
-        image = [self privateScreenshotImage];
-        if (!image) {
-            image = [self fallbackScreenshotImage];
+        @autoreleasepool {
+            image = [self privateScreenshotImage];
+            if (!image) {
+                image = [self fallbackScreenshotImage];
+            }
         }
     };
     if ([NSThread isMainThread]) {
@@ -589,11 +601,11 @@ - (UIImage *)pointSizedImageFromImage:(UIImage *)image {
 }
 
 - (NSData *)JPEGDataForImage:(UIImage *)image maxBytes:(NSUInteger)maxBytes {
-    NSData *bestData = UIImageJPEGRepresentation(image, kMCPScreenshotInitialJPEGQuality);
+    NSData *bestData = MCPJPEGRepresentation(image, kMCPScreenshotInitialJPEGQuality);
     if (!bestData) return nil;
     if (bestData.length <= maxBytes) return bestData;
 
-    NSData *minimumData = UIImageJPEGRepresentation(image, kMCPScreenshotMinimumJPEGQuality);
+    NSData *minimumData = MCPJPEGRepresentation(image, kMCPScreenshotMinimumJPEGQuality);
     if (!minimumData) return bestData;
     if (minimumData.length > maxBytes) return minimumData;
 
@@ -605,7 +617,7 @@ - (NSData *)JPEGDataForImage:(UIImage *)image maxBytes:(NSUInteger)maxBytes {
     CGFloat high = kMCPScreenshotInitialJPEGQuality;
     for (NSInteger pass = 0; pass < kMCPScreenshotJPEGSearchPasses; pass++) {
         CGFloat quality = (low + high) / 2.0;
-        NSData *candidate = UIImageJPEGRepresentation(image, quality);
+        NSData *candidate = MCPJPEGRepresentation(image, quality);
         if (!candidate) break;
 
         if (candidate.length > maxBytes) {

diff --git a/prefs/Resources/Root.plist b/prefs/Resources/Root.plist
@@ -54,6 +54,36 @@
 			<key>placeholder</key>
 			<string>8090</string>
 		</dict>
+		<dict>
+			<key>cell</key>
+			<string>PSGroupCell</string>
+			<key>label</key>
+			<string>图像处理</string>
+			<key>footerText</key>
+			<string>限制截图和 OCR 同时处理的任务数。1 更节省内存，2 可提高高内存设备的吞吐量。修改后需要重启 SpringBoard。</string>
+		</dict>
+		<dict>
+			<key>cell</key>
+			<string>PSSegmentCell</string>
+			<key>default</key>
+			<integer>1</integer>
+			<key>defaults</key>
+			<string>com.witchan.ios-mcp.preferences</string>
+			<key>key</key>
+			<string>maxConcurrentScreenTasks</string>
+			<key>label</key>
+			<string>并发任务</string>
+			<key>validTitles</key>
+			<array>
+				<string>1</string>
+				<string>2</string>
+			</array>
+			<key>validValues</key>
+			<array>
+				<integer>1</integer>
+				<integer>2</integer>
+			</array>
+		</dict>
 		<dict>
 			<key>cell</key>
 			<string>PSGroupCell</string>