@midscene/shared 1.9.7 → 1.9.8-beta-20260618014851.0

package/dist/lib/env/types.js

@@ -36,6 +36,7 @@ __webpack_require__.d(__webpack_exports__, {
     MIDSCENE_ANDROID_IME_STRATEGY: ()=>MIDSCENE_ANDROID_IME_STRATEGY,
     MIDSCENE_CACHE: ()=>MIDSCENE_CACHE,
     MIDSCENE_CACHE_MAX_FILENAME_LENGTH: ()=>MIDSCENE_CACHE_MAX_FILENAME_LENGTH,
+    MIDSCENE_CHROME_PATH: ()=>MIDSCENE_CHROME_PATH,
     MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG: ()=>MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG,
     MIDSCENE_DEBUG_MODE: ()=>MIDSCENE_DEBUG_MODE,
     MIDSCENE_DEBUG_MODEL_PROFILE: ()=>MIDSCENE_DEBUG_MODEL_PROFILE,
@@ -60,9 +61,7 @@ __webpack_require__.d(__webpack_exports__, {
     MIDSCENE_IOS_SIMULATOR_UDID: ()=>MIDSCENE_IOS_SIMULATOR_UDID,
     MIDSCENE_LANGFUSE_DEBUG: ()=>MIDSCENE_LANGFUSE_DEBUG,
     MIDSCENE_LANGSMITH_DEBUG: ()=>MIDSCENE_LANGSMITH_DEBUG,
-    MIDSCENE_MCP_ANDROID_MODE: ()=>MIDSCENE_MCP_ANDROID_MODE,
     MIDSCENE_MCP_CHROME_PATH: ()=>MIDSCENE_MCP_CHROME_PATH,
-    MIDSCENE_MCP_USE_PUPPETEER_MODE: ()=>MIDSCENE_MCP_USE_PUPPETEER_MODE,
     MIDSCENE_MODEL_API_KEY: ()=>MIDSCENE_MODEL_API_KEY,
     MIDSCENE_MODEL_BASE_URL: ()=>MIDSCENE_MODEL_BASE_URL,
     MIDSCENE_MODEL_EXTRA_BODY_JSON: ()=>MIDSCENE_MODEL_EXTRA_BODY_JSON,
@@ -123,9 +122,8 @@ const MIDSCENE_DEBUG_MODEL_PROFILE = 'MIDSCENE_DEBUG_MODEL_PROFILE';
 const MIDSCENE_DEBUG_MODEL_RESPONSE = 'MIDSCENE_DEBUG_MODEL_RESPONSE';
 const MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG = 'MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG';
 const MIDSCENE_DEBUG_MODE = 'MIDSCENE_DEBUG_MODE';
-const MIDSCENE_MCP_USE_PUPPETEER_MODE = 'MIDSCENE_MCP_USE_PUPPETEER_MODE';
+const MIDSCENE_CHROME_PATH = 'MIDSCENE_CHROME_PATH';
 const MIDSCENE_MCP_CHROME_PATH = 'MIDSCENE_MCP_CHROME_PATH';
-const MIDSCENE_MCP_ANDROID_MODE = 'MIDSCENE_MCP_ANDROID_MODE';
 const DOCKER_CONTAINER = 'DOCKER_CONTAINER';
 const MIDSCENE_LANGSMITH_DEBUG = 'MIDSCENE_LANGSMITH_DEBUG';
 const MIDSCENE_LANGFUSE_DEBUG = 'MIDSCENE_LANGFUSE_DEBUG';
@@ -208,8 +206,6 @@ const BASIC_ENV_KEYS = [
 ];
 const BOOLEAN_ENV_KEYS = [
     MIDSCENE_CACHE,
-    MIDSCENE_MCP_USE_PUPPETEER_MODE,
-    MIDSCENE_MCP_ANDROID_MODE,
     MIDSCENE_LANGSMITH_DEBUG,
     MIDSCENE_LANGFUSE_DEBUG,
     MIDSCENE_REPORT_QUIET
@@ -228,6 +224,7 @@ const STRING_ENV_KEYS = [
     MIDSCENE_REPORT_TAG_NAME,
     MIDSCENE_PREFERRED_LANGUAGE,
     MATCH_BY_POSITION,
+    MIDSCENE_CHROME_PATH,
     MIDSCENE_MCP_CHROME_PATH,
     DOCKER_CONTAINER
 ];
@@ -338,6 +335,7 @@ exports.MIDSCENE_ADB_REMOTE_PORT = __webpack_exports__.MIDSCENE_ADB_REMOTE_PORT;
 exports.MIDSCENE_ANDROID_IME_STRATEGY = __webpack_exports__.MIDSCENE_ANDROID_IME_STRATEGY;
 exports.MIDSCENE_CACHE = __webpack_exports__.MIDSCENE_CACHE;
 exports.MIDSCENE_CACHE_MAX_FILENAME_LENGTH = __webpack_exports__.MIDSCENE_CACHE_MAX_FILENAME_LENGTH;
+exports.MIDSCENE_CHROME_PATH = __webpack_exports__.MIDSCENE_CHROME_PATH;
 exports.MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG = __webpack_exports__.MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG;
 exports.MIDSCENE_DEBUG_MODE = __webpack_exports__.MIDSCENE_DEBUG_MODE;
 exports.MIDSCENE_DEBUG_MODEL_PROFILE = __webpack_exports__.MIDSCENE_DEBUG_MODEL_PROFILE;
@@ -362,9 +360,7 @@ exports.MIDSCENE_IOS_DEVICE_UDID = __webpack_exports__.MIDSCENE_IOS_DEVICE_UDID;
 exports.MIDSCENE_IOS_SIMULATOR_UDID = __webpack_exports__.MIDSCENE_IOS_SIMULATOR_UDID;
 exports.MIDSCENE_LANGFUSE_DEBUG = __webpack_exports__.MIDSCENE_LANGFUSE_DEBUG;
 exports.MIDSCENE_LANGSMITH_DEBUG = __webpack_exports__.MIDSCENE_LANGSMITH_DEBUG;
-exports.MIDSCENE_MCP_ANDROID_MODE = __webpack_exports__.MIDSCENE_MCP_ANDROID_MODE;
 exports.MIDSCENE_MCP_CHROME_PATH = __webpack_exports__.MIDSCENE_MCP_CHROME_PATH;
-exports.MIDSCENE_MCP_USE_PUPPETEER_MODE = __webpack_exports__.MIDSCENE_MCP_USE_PUPPETEER_MODE;
 exports.MIDSCENE_MODEL_API_KEY = __webpack_exports__.MIDSCENE_MODEL_API_KEY;
 exports.MIDSCENE_MODEL_BASE_URL = __webpack_exports__.MIDSCENE_MODEL_BASE_URL;
 exports.MIDSCENE_MODEL_EXTRA_BODY_JSON = __webpack_exports__.MIDSCENE_MODEL_EXTRA_BODY_JSON;
@@ -430,6 +426,7 @@ for(var __rspack_i in __webpack_exports__)if (-1 === [
     "MIDSCENE_ANDROID_IME_STRATEGY",
     "MIDSCENE_CACHE",
     "MIDSCENE_CACHE_MAX_FILENAME_LENGTH",
+    "MIDSCENE_CHROME_PATH",
     "MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG",
     "MIDSCENE_DEBUG_MODE",
     "MIDSCENE_DEBUG_MODEL_PROFILE",
@@ -454,9 +451,7 @@ for(var __rspack_i in __webpack_exports__)if (-1 === [
     "MIDSCENE_IOS_SIMULATOR_UDID",
     "MIDSCENE_LANGFUSE_DEBUG",
     "MIDSCENE_LANGSMITH_DEBUG",
-    "MIDSCENE_MCP_ANDROID_MODE",
     "MIDSCENE_MCP_CHROME_PATH",
-    "MIDSCENE_MCP_USE_PUPPETEER_MODE",
     "MIDSCENE_MODEL_API_KEY",
     "MIDSCENE_MODEL_BASE_URL",
     "MIDSCENE_MODEL_EXTRA_BODY_JSON",

package/dist/lib/img/index.js

@@ -33,6 +33,7 @@ __webpack_require__.d(__webpack_exports__, {
     saveBase64Image: ()=>external_transform_js_namespaceObject.saveBase64Image,
     isValidImageBuffer: ()=>external_info_js_namespaceObject.isValidImageBuffer,
     validateScreenshotBuffer: ()=>external_info_js_namespaceObject.validateScreenshotBuffer,
+    normalizeScreenshotBase64: ()=>external_transform_js_namespaceObject.normalizeScreenshotBase64,
     preProcessImageUrl: ()=>external_transform_js_namespaceObject.preProcessImageUrl,
     annotateRects: ()=>external_box_select_js_namespaceObject.annotateRects,
     compositeElementInfoImg: ()=>external_box_select_js_namespaceObject.compositeElementInfoImg,
@@ -64,6 +65,7 @@ exports.isValidJPEGImageBuffer = __webpack_exports__.isValidJPEGImageBuffer;
 exports.isValidPNGImageBuffer = __webpack_exports__.isValidPNGImageBuffer;
 exports.localImg2Base64 = __webpack_exports__.localImg2Base64;
 exports.normalizeBase64Image = __webpack_exports__.normalizeBase64Image;
+exports.normalizeScreenshotBase64 = __webpack_exports__.normalizeScreenshotBase64;
 exports.paddingToMatchBlockByBase64 = __webpack_exports__.paddingToMatchBlockByBase64;
 exports.parseBase64 = __webpack_exports__.parseBase64;
 exports.preProcessImageUrl = __webpack_exports__.preProcessImageUrl;
@@ -88,6 +90,7 @@ for(var __rspack_i in __webpack_exports__)if (-1 === [
     "isValidPNGImageBuffer",
     "localImg2Base64",
     "normalizeBase64Image",
+    "normalizeScreenshotBase64",
     "paddingToMatchBlockByBase64",
     "parseBase64",
     "preProcessImageUrl",

package/dist/lib/img/transform.js

@@ -42,6 +42,7 @@ __webpack_require__.d(__webpack_exports__, {
     resizeAndConvertImgBuffer: ()=>resizeAndConvertImgBuffer,
     resizeImgBase64: ()=>resizeImgBase64,
     saveBase64Image: ()=>saveBase64Image,
+    normalizeScreenshotBase64: ()=>normalizeScreenshotBase64,
     preProcessImageUrl: ()=>preProcessImageUrl,
     cropByRect: ()=>cropByRect,
     scaleImage: ()=>scaleImage,
@@ -129,6 +130,8 @@ async function resizeAndConvertImgBuffer(inputFormat, inputData, newSize) {
 }
 const normalizeBase64Body = (body)=>body.replace(/\s/g, '');
 const base64ImageDataUrlPattern = /^data:image\/[a-zA-Z0-9.+-]+;base64,/i;
+const supportedScreenshotDataUriPattern = /^data:image\/(png|jpe?g);base64,([\s\S]*)$/i;
+const rawBase64BodyPattern = /^[A-Za-z0-9+/=\s]+$/;
 const inferBase64ImageFormat = (base64Body)=>{
     if (base64Body.startsWith('iVBORw0KGgo')) return 'png';
     return 'jpeg';
@@ -141,6 +144,21 @@ function detectImageMimeTypeFromBuffer(buffer) {
     if (buffer.length >= 2 && 0x42 === buffer[0] && 0x4d === buffer[1]) return 'image/bmp';
 }
 const createImgBase64ByFormat = (format, body)=>`data:image/${format};base64,${normalizeBase64Body(body)}`;
+const normalizeScreenshotBase64 = (base64, options)=>{
+    const label = options?.label ?? 'screenshot base64';
+    const trimmedBase64 = base64.trim();
+    if (!trimmedBase64) throw new Error(`${label} cannot be empty`);
+    const dataUriMatch = trimmedBase64.match(supportedScreenshotDataUriPattern);
+    if (dataUriMatch) {
+        const imageFormat = 'jpg' === dataUriMatch[1].toLowerCase() ? 'jpeg' : dataUriMatch[1].toLowerCase();
+        const body = dataUriMatch[2];
+        if (!normalizeBase64Body(body)) throw new Error(`${label} cannot be empty`);
+        return createImgBase64ByFormat(imageFormat, body);
+    }
+    if (trimmedBase64.startsWith('data:')) throw new Error(`${label} must be a PNG/JPEG data URI or raw PNG base64 string`);
+    if (!rawBase64BodyPattern.test(trimmedBase64)) throw new Error(`${label} must be a PNG/JPEG data URI or raw PNG base64 string`);
+    return createImgBase64ByFormat('png', trimmedBase64);
+};
 const normalizeBase64Image = (base64)=>{
     const trimmedBase64 = base64.trim();
     if (base64ImageDataUrlPattern.test(trimmedBase64)) return trimmedBase64;
@@ -358,6 +376,7 @@ exports.inferBase64ImageFormat = __webpack_exports__.inferBase64ImageFormat;
 exports.localImg2Base64 = __webpack_exports__.localImg2Base64;
 exports.normalizeBase64Body = __webpack_exports__.normalizeBase64Body;
 exports.normalizeBase64Image = __webpack_exports__.normalizeBase64Image;
+exports.normalizeScreenshotBase64 = __webpack_exports__.normalizeScreenshotBase64;
 exports.paddingToMatchBlock = __webpack_exports__.paddingToMatchBlock;
 exports.paddingToMatchBlockByBase64 = __webpack_exports__.paddingToMatchBlockByBase64;
 exports.parseBase64 = __webpack_exports__.parseBase64;
@@ -377,6 +396,7 @@ for(var __rspack_i in __webpack_exports__)if (-1 === [
     "localImg2Base64",
     "normalizeBase64Body",
     "normalizeBase64Image",
+    "normalizeScreenshotBase64",
     "paddingToMatchBlock",
     "paddingToMatchBlockByBase64",
     "parseBase64",

package/dist/lib/utils.js

@@ -24,20 +24,19 @@ var __webpack_require__ = {};
 var __webpack_exports__ = {};
 __webpack_require__.r(__webpack_exports__);
 __webpack_require__.d(__webpack_exports__, {
-    uuid: ()=>uuid,
-    isPlainObject: ()=>isPlainObject,
-    ifInBrowser: ()=>ifInBrowser,
-    repeat: ()=>repeat,
-    mergeAndNormalizeAppNameMapping: ()=>mergeAndNormalizeAppNameMapping,
-    ifInNode: ()=>ifInNode,
-    assert: ()=>assert,
     escapeScriptTag: ()=>escapeScriptTag,
     logMsg: ()=>logMsg,
+    isPlainObject: ()=>isPlainObject,
     replaceIllegalPathCharsAndSpace: ()=>replaceIllegalPathCharsAndSpace,
+    uuid: ()=>uuid,
+    ifInBrowser: ()=>ifInBrowser,
     generateHashId: ()=>generateHashId,
     normalizeForComparison: ()=>normalizeForComparison,
+    mergeAndNormalizeAppNameMapping: ()=>mergeAndNormalizeAppNameMapping,
+    ifInNode: ()=>ifInNode,
     ifInWorker: ()=>ifInWorker,
-    setIsMcp: ()=>setIsMcp,
+    repeat: ()=>repeat,
+    assert: ()=>assert,
     antiEscapeScriptTag: ()=>antiEscapeScriptTag
 });
 const external_js_sha256_namespaceObject = require("js-sha256");
@@ -76,12 +75,8 @@ function generateHashId(rect, content = '') {
 function assert(condition, message) {
     if (!condition) throw new Error(message || 'Assertion failed');
 }
-let isMcp = false;
-function setIsMcp(value) {
-    isMcp = value;
-}
 function logMsg(...message) {
-    if (!isMcp) console.log(...message);
+    console.log(...message);
 }
 async function repeat(times, fn) {
     for(let i = 0; i < times; i++)await fn(i);
@@ -124,7 +119,6 @@ exports.mergeAndNormalizeAppNameMapping = __webpack_exports__.mergeAndNormalizeA
 exports.normalizeForComparison = __webpack_exports__.normalizeForComparison;
 exports.repeat = __webpack_exports__.repeat;
 exports.replaceIllegalPathCharsAndSpace = __webpack_exports__.replaceIllegalPathCharsAndSpace;
-exports.setIsMcp = __webpack_exports__.setIsMcp;
 exports.uuid = __webpack_exports__.uuid;
 for(var __rspack_i in __webpack_exports__)if (-1 === [
     "antiEscapeScriptTag",
@@ -140,7 +134,6 @@ for(var __rspack_i in __webpack_exports__)if (-1 === [
     "normalizeForComparison",
     "repeat",
     "replaceIllegalPathCharsAndSpace",
-    "setIsMcp",
     "uuid"
 ].indexOf(__rspack_i)) exports[__rspack_i] = __webpack_exports__[__rspack_i];
 Object.defineProperty(exports, '__esModule', {

package/dist/types/agent-tools/agent-behavior-init-args.d.ts

@@ -0,0 +1,17 @@
+import { z } from 'zod';
+export interface AgentBehaviorInitArgs {
+    aiActContext?: string;
+    aiActionContext?: string;
+    replanningCycleLimit?: number;
+    waitAfterAction?: number;
+    screenshotShrinkFactor?: number;
+}
+export declare const agentBehaviorInitArgShape: {
+    aiActContext: z.ZodOptional<z.ZodString>;
+    replanningCycleLimit: z.ZodOptional<z.ZodNumber>;
+    waitAfterAction: z.ZodOptional<z.ZodNumber>;
+    screenshotShrinkFactor: z.ZodOptional<z.ZodNumber>;
+};
+export declare function extractAgentBehaviorInitArgs(extracted: Partial<AgentBehaviorInitArgs> | undefined): AgentBehaviorInitArgs | undefined;
+export declare function getAgentInitArgsSignature(initArgs: object | undefined): string | undefined;
+export declare function shouldRebuildAgentForInitArgs(currentSignature: string | undefined, nextSignature: string | undefined): boolean;

package/dist/types/{mcp → agent-tools}/base-tools.d.ts

@@ -1,4 +1,3 @@
-import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
 import type { z } from 'zod';
 import { type CliReportSession } from './cli-report-session';
 import { type ToolDefaults } from './tool-defaults';
@@ -11,11 +10,11 @@ import type { BaseAgent, BaseDevice, IMidsceneTools, ToolCliMetadata, ToolDefini
 export interface InitArgSpec<TInitParam> {
     /** Arg namespace, e.g. `android`, `ios`. */
     namespace: string;
-    /** Zod shape describing the init args. Field names drive the MCP schema. */
+    /** Zod shape describing the init args. Field names drive the tool schema. */
     shape: Record<string, z.ZodTypeAny>;
     /**
      * Optional CLI presentation hints. These affect `--help` output for
-     * single-platform CLIs but do not alter MCP/YAML protocol keys.
+     * single-platform CLIs but do not alter YAML protocol keys.
      */
     cli?: {
         /** Prefer bare `--device-id`-style options in platform CLI help output. */
@@ -30,24 +29,23 @@ export interface InitArgSpec<TInitParam> {
     adapt?: (extracted: Record<string, unknown> | undefined) => TInitParam | undefined;
 }
 /**
- * Base class for platform-specific MCP tools.
+ * Base class for platform-specific Midscene tools.
  * @typeParam TAgent - Platform-specific agent type.
  * @typeParam TInitParam - Platform-specific init parameter consumed by
  *   `ensureAgent`. Defaults to `undefined` for platforms that take no args.
  */
 export declare abstract class BaseMidsceneTools<TAgent extends BaseAgent = BaseAgent, TInitParam = unknown> implements IMidsceneTools {
-    protected mcpServer?: McpServer;
     protected agent?: TAgent;
     protected toolDefinitions: ToolDefinition[];
     /**
      * Default options injected into every generated tool call (e.g. forced deep
-     * locate / deep think). Set from server/CLI behavior flags before
+     * locate / deep think). Set from startup/CLI behavior flags before
      * `initTools()` so they are baked into the generated tool handlers.
      * See https://github.com/web-infra-dev/midscene/issues/2446.
      */
     protected toolDefaults: ToolDefaults;
     /**
-     * Declarative init-arg spec. Subclasses that accept CLI/MCP init args should
+     * Declarative init-arg spec. Subclasses that accept CLI init args should
      * set this once and get `extractAgentInitParam` / `sanitizeToolArgs` /
      * `getAgentInitArgSchema` auto-implemented.
      *
@@ -66,7 +64,7 @@ export declare abstract class BaseMidsceneTools<TAgent extends BaseAgent = BaseA
     protected abstract ensureAgent(initParam?: TInitParam): Promise<TAgent>;
     private getInitArgKeys;
     /**
-     * Extract a platform-specific agent init parameter from CLI/MCP tool args.
+     * Extract a platform-specific agent init parameter from CLI tool args.
      */
     protected extractAgentInitParam(args: Record<string, unknown>): TInitParam | undefined;
     /**
@@ -82,7 +80,7 @@ export declare abstract class BaseMidsceneTools<TAgent extends BaseAgent = BaseA
      * show ergonomic bare flags while the underlying schema stays namespaced.
      * When `preferBareKeys` is enabled, single-platform CLIs only accept the
      * bare spellings; namespaced dotted spellings remain available through the
-     * MCP/YAML schema instead of the platform CLI surface.
+     * YAML schema instead of the platform CLI surface.
      */
     protected getAgentInitArgCliMetadata(): ToolCliMetadata | undefined;
     /**
@@ -109,10 +107,6 @@ export declare abstract class BaseMidsceneTools<TAgent extends BaseAgent = BaseA
      * 2. Create temporary device instance to read actionSpace (always succeeds)
      */
     initTools(): Promise<void>;
-    /**
-     * Attach to MCP server and register all tools
-     */
-    attachToServer(server: McpServer): void;
     /**
      * Cleanup method - destroy agent and release resources
      */

package/dist/types/{mcp → agent-tools}/index.d.ts

@@ -1,10 +1,8 @@
-export * from './base-server';
 export * from './base-tools';
 export * from './tool-defaults';
+export * from './agent-behavior-init-args';
 export * from './init-arg-utils';
 export * from './error-formatter';
 export * from './tool-generator';
 export * from './types';
-export * from './inject-report-html-plugin';
-export * from './launcher-helper';
 export * from './chrome-path';

package/dist/types/{mcp → agent-tools}/init-arg-utils.d.ts

@@ -3,11 +3,11 @@ import type { ToolSchema } from './types';
 export declare function extractNamespacedArgs<TFieldName extends string, TArgs extends Record<string, unknown> = Record<string, unknown>>(args: Record<string, unknown>, namespace: string, keys: readonly TFieldName[]): TArgs | undefined;
 export declare function sanitizeNamespacedArgs(args: Record<string, unknown>, namespace: string, keys: readonly string[]): Record<string, unknown>;
 /**
- * Build a flat MCP tool schema whose keys are dotted `"<namespace>.<field>"`.
+ * Build a flat tool schema whose keys are dotted `"<namespace>.<field>"`.
  *
  * We intentionally stay flat (rather than `{ namespace: z.object({...}) }`) so
- * that CLI (`--android.device-id`), MCP clients, and `--help` output all share
- * the same spelling. `readNamespacedArg` understands all three input shapes:
+ * that CLI (`--android.device-id`) and `--help` output share the same spelling.
+ * `readNamespacedArg` understands all three input shapes:
  * nested namespace object, dotted flat key, and bare key fallback.
  */
 export declare function createNamespacedInitArgSchema(namespace: string, shape: Record<string, z.ZodTypeAny>): ToolSchema;

package/dist/types/{mcp → agent-tools}/tool-defaults.d.ts

@@ -1,10 +1,10 @@
 /**
  * Unified, declarative mechanism for "force a default option on every tool
- * call" behaviors exposed by MCP servers and the device / Agent Skill CLIs.
+ * call" behaviors exposed by device and Agent Skill CLIs.
  *
  * Adding a new behavior flag (e.g. `--deep-search`) is a one-line change to
  * {@link TOOL_BEHAVIOR_FLAGS}: declare which default-option "bag" it fills.
- * The tool generator, servers, tools managers and CLI parsing are all generic
+ * The tool generator, tools managers and CLI parsing are all generic
  * over {@link ToolDefaults} and never need to learn about individual flags.
  *
  * See https://github.com/web-infra-dev/midscene/issues/2446.
@@ -50,13 +50,12 @@ export declare function resolveToolDefaults(isEnabled: (cli: string) => boolean)
  *
  * Behavior flags (e.g. `--deep-locate`) are global: they may appear anywhere
  * in argv and are not tied to a specific sub-command. They are recognized by
- * exact kebab-case match — the same surface the MCP `parseArgs` config exposes
- * — and removed so a strict per-command parser never sees them. Every other
+ * exact kebab-case match and removed so a strict per-command parser never sees them. Every other
  * token is returned untouched and in order for that per-command parser.
  *
  * This is the single place that knows how a behavior flag looks on the command
- * line; both the device / Agent Skill CLI and the MCP launch path resolve their
- * defaults from {@link TOOL_BEHAVIOR_FLAGS} through here / {@link resolveToolDefaults}.
+ * line; the device / Agent Skill CLI resolves defaults from
+ * {@link TOOL_BEHAVIOR_FLAGS} through here / {@link resolveToolDefaults}.
  */
 export declare function stripBehaviorFlags(argv: readonly string[]): {
     rawArgs: string[];

package/dist/types/{mcp → agent-tools}/tool-generator.d.ts

@@ -3,7 +3,7 @@ import type { ActionSpaceItem, BaseAgent, ToolCliMetadata, ToolDefinition, ToolS
 import { composeUserPrompt } from './user-prompt';
 export { composeUserPrompt };
 /**
- * Converts DeviceAction from actionSpace into MCP ToolDefinition
+ * Converts DeviceAction from actionSpace into ToolDefinition.
  * This is the core logic that removes need for hardcoded tool definitions
  */
 export declare function generateToolsFromActionSpace(actionSpace: ActionSpaceItem[], getAgent: (args?: Record<string, unknown>) => Promise<BaseAgent>, sanitizeArgs?: (args: Record<string, unknown>) => Record<string, unknown>, initArgSchema?: ToolSchema, initArgCliMetadata?: ToolCliMetadata, toolDefaults?: ToolDefaults): ToolDefinition[];

package/dist/types/{mcp → agent-tools}/types.d.ts

@@ -1,4 +1,3 @@
-import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
 import type { z } from 'zod';
 import type { ToolDefaults } from './tool-defaults';
 /**
@@ -7,7 +6,7 @@ import type { ToolDefaults } from './tool-defaults';
 export declare const defaultAppLoadingTimeoutMs = 10000;
 export declare const defaultAppLoadingCheckIntervalMs = 2000;
 /**
- * Content item types for tool results (MCP compatible)
+ * Content item types for tool results.
  */
 export type ToolResultContent = {
     type: 'text';
@@ -33,7 +32,7 @@ export type ToolResultContent = {
     };
 };
 /**
- * Result type for tool execution (MCP compatible)
+ * Result type for tool execution.
  */
 export interface ToolResult {
     [x: string]: unknown;
@@ -58,7 +57,7 @@ export interface ToolCliMetadata {
     options?: Record<string, ToolCliOption>;
 }
 /**
- * Tool definition for MCP server
+ * Tool definition for Midscene CLI and Skill surfaces.
  */
 export interface ToolDefinition<T = Record<string, unknown>> {
     name: string;
@@ -67,9 +66,6 @@ export interface ToolDefinition<T = Record<string, unknown>> {
     handler: ToolHandler<T>;
     cli?: ToolCliMetadata;
 }
-/**
- * Tool type for mcpKitForAgent return value
- */
 export type Tool = ToolDefinition;
 /**
  * Action space item definition
@@ -97,6 +93,21 @@ export type UserPromptLike = string | {
     }>;
     convertHttpImage2Base64?: boolean;
 };
+export interface RecordToReportScreenshot {
+    /**
+     * PNG/JPEG data URI, or raw PNG base64 body.
+     */
+    base64: string;
+    description?: string;
+}
+export interface RecordToReportOptions {
+    content?: string;
+    /**
+     * @deprecated Use `screenshots: [{ base64 }]` instead.
+     */
+    screenshotBase64?: string;
+    screenshots?: RecordToReportScreenshot[];
+}
 /**
  * Base agent interface
  * Represents a platform-specific agent (Android, iOS, Web)
@@ -108,10 +119,7 @@ export interface BaseAgent {
     page?: {
         screenshotBase64(): Promise<string>;
     };
-    recordToReport?: (title?: string, opt?: {
-        content?: string;
-        screenshotBase64?: string;
-    }) => Promise<void>;
+    recordToReport?: (title?: string, opt?: RecordToReportOptions) => Promise<void>;
     callActionInActionSpace?: (actionName: string, params?: unknown) => Promise<unknown>;
     aiAction?: (description: string, params?: Record<string, unknown>) => Promise<unknown>;
     aiWaitFor?: (assertion: string, options: Record<string, unknown>) => Promise<unknown>;
@@ -125,10 +133,9 @@ export interface BaseDevice {
     destroy?(): Promise<void>;
 }
 /**
- * Interface for platform-specific MCP tools manager
+ * Interface for platform-specific tools manager.
  */
 export interface IMidsceneTools {
-    attachToServer(server: McpServer): void;
     initTools(): Promise<void>;
     destroy?(): Promise<void>;
     setToolDefaults?(toolDefaults: ToolDefaults): void;

package/dist/types/cli/cli-args.d.ts

@@ -1,4 +1,4 @@
-import type { ToolCliOption, ToolDefinition } from '../mcp/types';
+import type { ToolCliOption, ToolDefinition } from '../agent-tools/types';
 export declare function parseValue(raw: string): unknown;
 export declare function parseCliArgs(args: string[]): Record<string, unknown>;
 export declare function getCliOptionDisplay(key: string, cliOption?: ToolCliOption): {

package/dist/types/cli/cli-runner.d.ts

@@ -1,5 +1,5 @@
-import type { BaseMidsceneTools } from '../mcp/base-tools';
-import type { ToolDefinition } from '../mcp/types';
+import type { BaseMidsceneTools } from '../agent-tools/base-tools';
+import type { ToolDefinition } from '../agent-tools/types';
 export interface CLIExtraCommand {
     name: string;
     def: ToolDefinition;

package/dist/types/env/types.d.ts

@@ -5,9 +5,11 @@ export declare const MIDSCENE_DEBUG_MODEL_PROFILE = "MIDSCENE_DEBUG_MODEL_PROFIL
 export declare const MIDSCENE_DEBUG_MODEL_RESPONSE = "MIDSCENE_DEBUG_MODEL_RESPONSE";
 export declare const MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG = "MIDSCENE_DANGEROUSLY_PRINT_ALL_CONFIG";
 export declare const MIDSCENE_DEBUG_MODE = "MIDSCENE_DEBUG_MODE";
-export declare const MIDSCENE_MCP_USE_PUPPETEER_MODE = "MIDSCENE_MCP_USE_PUPPETEER_MODE";
+export declare const MIDSCENE_CHROME_PATH = "MIDSCENE_CHROME_PATH";
+/**
+ * @deprecated Use MIDSCENE_CHROME_PATH instead. This is kept for backward compatibility.
+ */
 export declare const MIDSCENE_MCP_CHROME_PATH = "MIDSCENE_MCP_CHROME_PATH";
-export declare const MIDSCENE_MCP_ANDROID_MODE = "MIDSCENE_MCP_ANDROID_MODE";
 export declare const DOCKER_CONTAINER = "DOCKER_CONTAINER";
 export declare const MIDSCENE_LANGSMITH_DEBUG = "MIDSCENE_LANGSMITH_DEBUG";
 export declare const MIDSCENE_LANGFUSE_DEBUG = "MIDSCENE_LANGFUSE_DEBUG";
@@ -104,22 +106,22 @@ export declare const UNUSED_ENV_KEYS: string[];
  * can not be override by overrideAIConfig
  */
 export declare const BASIC_ENV_KEYS: readonly ["MIDSCENE_DEBUG_MODE", "MIDSCENE_DEBUG_MODEL_PROFILE", "MIDSCENE_DEBUG_MODEL_RESPONSE", "MIDSCENE_RUN_DIR"];
-export declare const BOOLEAN_ENV_KEYS: readonly ["MIDSCENE_CACHE", "MIDSCENE_MCP_USE_PUPPETEER_MODE", "MIDSCENE_MCP_ANDROID_MODE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_LANGFUSE_DEBUG", "MIDSCENE_REPORT_QUIET"];
+export declare const BOOLEAN_ENV_KEYS: readonly ["MIDSCENE_CACHE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_LANGFUSE_DEBUG", "MIDSCENE_REPORT_QUIET"];
 export declare const NUMBER_ENV_KEYS: readonly ["MIDSCENE_CACHE_MAX_FILENAME_LENGTH", "MIDSCENE_REPLANNING_CYCLE_LIMIT"];
-export declare const STRING_ENV_KEYS: readonly ["MIDSCENE_ADB_PATH", "MIDSCENE_ADB_REMOTE_HOST", "MIDSCENE_ADB_REMOTE_PORT", "MIDSCENE_ANDROID_IME_STRATEGY", "MIDSCENE_IOS_DEVICE_UDID", "MIDSCENE_IOS_SIMULATOR_UDID", "MIDSCENE_REPORT_TAG_NAME", "MIDSCENE_PREFERRED_LANGUAGE", "MATCH_BY_POSITION", "MIDSCENE_MCP_CHROME_PATH", "DOCKER_CONTAINER"];
+export declare const STRING_ENV_KEYS: readonly ["MIDSCENE_ADB_PATH", "MIDSCENE_ADB_REMOTE_HOST", "MIDSCENE_ADB_REMOTE_PORT", "MIDSCENE_ANDROID_IME_STRATEGY", "MIDSCENE_IOS_DEVICE_UDID", "MIDSCENE_IOS_SIMULATOR_UDID", "MIDSCENE_REPORT_TAG_NAME", "MIDSCENE_PREFERRED_LANGUAGE", "MATCH_BY_POSITION", "MIDSCENE_CHROME_PATH", "MIDSCENE_MCP_CHROME_PATH", "DOCKER_CONTAINER"];
 /**
  * Non model related env keys, used for globally controlling the behavior of midscene
  * Can not be override by agent.modelConfig but can be override by overrideAIConfig
  * Can be access at any time
  */
-export declare const GLOBAL_ENV_KEYS: readonly ["MIDSCENE_CACHE", "MIDSCENE_MCP_USE_PUPPETEER_MODE", "MIDSCENE_MCP_ANDROID_MODE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_LANGFUSE_DEBUG", "MIDSCENE_REPORT_QUIET", "MIDSCENE_CACHE_MAX_FILENAME_LENGTH", "MIDSCENE_REPLANNING_CYCLE_LIMIT", "MIDSCENE_ADB_PATH", "MIDSCENE_ADB_REMOTE_HOST", "MIDSCENE_ADB_REMOTE_PORT", "MIDSCENE_ANDROID_IME_STRATEGY", "MIDSCENE_IOS_DEVICE_UDID", "MIDSCENE_IOS_SIMULATOR_UDID", "MIDSCENE_REPORT_TAG_NAME", "MIDSCENE_PREFERRED_LANGUAGE", "MATCH_BY_POSITION", "MIDSCENE_MCP_CHROME_PATH", "DOCKER_CONTAINER"];
+export declare const GLOBAL_ENV_KEYS: readonly ["MIDSCENE_CACHE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_LANGFUSE_DEBUG", "MIDSCENE_REPORT_QUIET", "MIDSCENE_CACHE_MAX_FILENAME_LENGTH", "MIDSCENE_REPLANNING_CYCLE_LIMIT", "MIDSCENE_ADB_PATH", "MIDSCENE_ADB_REMOTE_HOST", "MIDSCENE_ADB_REMOTE_PORT", "MIDSCENE_ANDROID_IME_STRATEGY", "MIDSCENE_IOS_DEVICE_UDID", "MIDSCENE_IOS_SIMULATOR_UDID", "MIDSCENE_REPORT_TAG_NAME", "MIDSCENE_PREFERRED_LANGUAGE", "MATCH_BY_POSITION", "MIDSCENE_CHROME_PATH", "MIDSCENE_MCP_CHROME_PATH", "DOCKER_CONTAINER"];
 /**
  * Model related eve keys, used for declare which model to use.
  * Can be override by both agent.modelConfig and overrideAIConfig
  * Can only be access after agent.constructor
  */
 export declare const MODEL_ENV_KEYS: readonly ["MIDSCENE_MODEL_NAME", "MIDSCENE_MODEL_INIT_CONFIG_JSON", "MIDSCENE_MODEL_EXTRA_BODY_JSON", "MIDSCENE_MODEL_API_KEY", "MIDSCENE_MODEL_BASE_URL", "MIDSCENE_MODEL_SOCKS_PROXY", "MIDSCENE_MODEL_HTTP_PROXY", "MIDSCENE_MODEL_TIMEOUT", "MIDSCENE_MODEL_TEMPERATURE", "MIDSCENE_MODEL_RETRY_COUNT", "MIDSCENE_MODEL_RETRY_INTERVAL", "MIDSCENE_MODEL_REASONING_EFFORT", "MIDSCENE_MODEL_REASONING_ENABLED", "MIDSCENE_MODEL_REASONING_BUDGET", "MIDSCENE_USE_VLM_UI_TARS", "MIDSCENE_USE_QWEN_VL", "MIDSCENE_USE_QWEN3_VL", "MIDSCENE_USE_DOUBAO_VISION", "MIDSCENE_USE_GEMINI", "MIDSCENE_USE_VL_MODEL", "OPENAI_API_KEY", "OPENAI_BASE_URL", "MIDSCENE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_OPENAI_HTTP_PROXY", "MIDSCENE_OPENAI_SOCKS_PROXY", "MIDSCENE_INSIGHT_MODEL_NAME", "MIDSCENE_INSIGHT_MODEL_SOCKS_PROXY", "MIDSCENE_INSIGHT_MODEL_HTTP_PROXY", "MIDSCENE_INSIGHT_MODEL_BASE_URL", "MIDSCENE_INSIGHT_MODEL_API_KEY", "MIDSCENE_INSIGHT_MODEL_INIT_CONFIG_JSON", "MIDSCENE_INSIGHT_MODEL_EXTRA_BODY_JSON", "MIDSCENE_INSIGHT_MODEL_TIMEOUT", "MIDSCENE_INSIGHT_MODEL_TEMPERATURE", "MIDSCENE_INSIGHT_MODEL_RETRY_COUNT", "MIDSCENE_INSIGHT_MODEL_RETRY_INTERVAL", "MIDSCENE_INSIGHT_MODEL_FAMILY", "MIDSCENE_INSIGHT_MODEL_REASONING_EFFORT", "MIDSCENE_INSIGHT_MODEL_REASONING_ENABLED", "MIDSCENE_INSIGHT_MODEL_REASONING_BUDGET", "MIDSCENE_PLANNING_MODEL_NAME", "MIDSCENE_PLANNING_MODEL_SOCKS_PROXY", "MIDSCENE_PLANNING_MODEL_HTTP_PROXY", "MIDSCENE_PLANNING_MODEL_BASE_URL", "MIDSCENE_PLANNING_MODEL_API_KEY", "MIDSCENE_PLANNING_MODEL_INIT_CONFIG_JSON", "MIDSCENE_PLANNING_MODEL_EXTRA_BODY_JSON", "MIDSCENE_PLANNING_MODEL_TIMEOUT", "MIDSCENE_PLANNING_MODEL_TEMPERATURE", "MIDSCENE_PLANNING_MODEL_RETRY_COUNT", "MIDSCENE_PLANNING_MODEL_RETRY_INTERVAL", "MIDSCENE_PLANNING_MODEL_FAMILY", "MIDSCENE_PLANNING_MODEL_REASONING_EFFORT", "MIDSCENE_PLANNING_MODEL_REASONING_ENABLED", "MIDSCENE_PLANNING_MODEL_REASONING_BUDGET", "MIDSCENE_MODEL_FAMILY"];
-export declare const ALL_ENV_KEYS: readonly [...string[], "MIDSCENE_DEBUG_MODE", "MIDSCENE_DEBUG_MODEL_PROFILE", "MIDSCENE_DEBUG_MODEL_RESPONSE", "MIDSCENE_RUN_DIR", "MIDSCENE_CACHE", "MIDSCENE_MCP_USE_PUPPETEER_MODE", "MIDSCENE_MCP_ANDROID_MODE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_LANGFUSE_DEBUG", "MIDSCENE_REPORT_QUIET", "MIDSCENE_CACHE_MAX_FILENAME_LENGTH", "MIDSCENE_REPLANNING_CYCLE_LIMIT", "MIDSCENE_ADB_PATH", "MIDSCENE_ADB_REMOTE_HOST", "MIDSCENE_ADB_REMOTE_PORT", "MIDSCENE_ANDROID_IME_STRATEGY", "MIDSCENE_IOS_DEVICE_UDID", "MIDSCENE_IOS_SIMULATOR_UDID", "MIDSCENE_REPORT_TAG_NAME", "MIDSCENE_PREFERRED_LANGUAGE", "MATCH_BY_POSITION", "MIDSCENE_MCP_CHROME_PATH", "DOCKER_CONTAINER", "MIDSCENE_MODEL_NAME", "MIDSCENE_MODEL_INIT_CONFIG_JSON", "MIDSCENE_MODEL_EXTRA_BODY_JSON", "MIDSCENE_MODEL_API_KEY", "MIDSCENE_MODEL_BASE_URL", "MIDSCENE_MODEL_SOCKS_PROXY", "MIDSCENE_MODEL_HTTP_PROXY", "MIDSCENE_MODEL_TIMEOUT", "MIDSCENE_MODEL_TEMPERATURE", "MIDSCENE_MODEL_RETRY_COUNT", "MIDSCENE_MODEL_RETRY_INTERVAL", "MIDSCENE_MODEL_REASONING_EFFORT", "MIDSCENE_MODEL_REASONING_ENABLED", "MIDSCENE_MODEL_REASONING_BUDGET", "MIDSCENE_USE_VLM_UI_TARS", "MIDSCENE_USE_QWEN_VL", "MIDSCENE_USE_QWEN3_VL", "MIDSCENE_USE_DOUBAO_VISION", "MIDSCENE_USE_GEMINI", "MIDSCENE_USE_VL_MODEL", "OPENAI_API_KEY", "OPENAI_BASE_URL", "MIDSCENE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_OPENAI_HTTP_PROXY", "MIDSCENE_OPENAI_SOCKS_PROXY", "MIDSCENE_INSIGHT_MODEL_NAME", "MIDSCENE_INSIGHT_MODEL_SOCKS_PROXY", "MIDSCENE_INSIGHT_MODEL_HTTP_PROXY", "MIDSCENE_INSIGHT_MODEL_BASE_URL", "MIDSCENE_INSIGHT_MODEL_API_KEY", "MIDSCENE_INSIGHT_MODEL_INIT_CONFIG_JSON", "MIDSCENE_INSIGHT_MODEL_EXTRA_BODY_JSON", "MIDSCENE_INSIGHT_MODEL_TIMEOUT", "MIDSCENE_INSIGHT_MODEL_TEMPERATURE", "MIDSCENE_INSIGHT_MODEL_RETRY_COUNT", "MIDSCENE_INSIGHT_MODEL_RETRY_INTERVAL", "MIDSCENE_INSIGHT_MODEL_FAMILY", "MIDSCENE_INSIGHT_MODEL_REASONING_EFFORT", "MIDSCENE_INSIGHT_MODEL_REASONING_ENABLED", "MIDSCENE_INSIGHT_MODEL_REASONING_BUDGET", "MIDSCENE_PLANNING_MODEL_NAME", "MIDSCENE_PLANNING_MODEL_SOCKS_PROXY", "MIDSCENE_PLANNING_MODEL_HTTP_PROXY", "MIDSCENE_PLANNING_MODEL_BASE_URL", "MIDSCENE_PLANNING_MODEL_API_KEY", "MIDSCENE_PLANNING_MODEL_INIT_CONFIG_JSON", "MIDSCENE_PLANNING_MODEL_EXTRA_BODY_JSON", "MIDSCENE_PLANNING_MODEL_TIMEOUT", "MIDSCENE_PLANNING_MODEL_TEMPERATURE", "MIDSCENE_PLANNING_MODEL_RETRY_COUNT", "MIDSCENE_PLANNING_MODEL_RETRY_INTERVAL", "MIDSCENE_PLANNING_MODEL_FAMILY", "MIDSCENE_PLANNING_MODEL_REASONING_EFFORT", "MIDSCENE_PLANNING_MODEL_REASONING_ENABLED", "MIDSCENE_PLANNING_MODEL_REASONING_BUDGET", "MIDSCENE_MODEL_FAMILY"];
+export declare const ALL_ENV_KEYS: readonly [...string[], "MIDSCENE_DEBUG_MODE", "MIDSCENE_DEBUG_MODEL_PROFILE", "MIDSCENE_DEBUG_MODEL_RESPONSE", "MIDSCENE_RUN_DIR", "MIDSCENE_CACHE", "MIDSCENE_LANGSMITH_DEBUG", "MIDSCENE_LANGFUSE_DEBUG", "MIDSCENE_REPORT_QUIET", "MIDSCENE_CACHE_MAX_FILENAME_LENGTH", "MIDSCENE_REPLANNING_CYCLE_LIMIT", "MIDSCENE_ADB_PATH", "MIDSCENE_ADB_REMOTE_HOST", "MIDSCENE_ADB_REMOTE_PORT", "MIDSCENE_ANDROID_IME_STRATEGY", "MIDSCENE_IOS_DEVICE_UDID", "MIDSCENE_IOS_SIMULATOR_UDID", "MIDSCENE_REPORT_TAG_NAME", "MIDSCENE_PREFERRED_LANGUAGE", "MATCH_BY_POSITION", "MIDSCENE_CHROME_PATH", "MIDSCENE_MCP_CHROME_PATH", "DOCKER_CONTAINER", "MIDSCENE_MODEL_NAME", "MIDSCENE_MODEL_INIT_CONFIG_JSON", "MIDSCENE_MODEL_EXTRA_BODY_JSON", "MIDSCENE_MODEL_API_KEY", "MIDSCENE_MODEL_BASE_URL", "MIDSCENE_MODEL_SOCKS_PROXY", "MIDSCENE_MODEL_HTTP_PROXY", "MIDSCENE_MODEL_TIMEOUT", "MIDSCENE_MODEL_TEMPERATURE", "MIDSCENE_MODEL_RETRY_COUNT", "MIDSCENE_MODEL_RETRY_INTERVAL", "MIDSCENE_MODEL_REASONING_EFFORT", "MIDSCENE_MODEL_REASONING_ENABLED", "MIDSCENE_MODEL_REASONING_BUDGET", "MIDSCENE_USE_VLM_UI_TARS", "MIDSCENE_USE_QWEN_VL", "MIDSCENE_USE_QWEN3_VL", "MIDSCENE_USE_DOUBAO_VISION", "MIDSCENE_USE_GEMINI", "MIDSCENE_USE_VL_MODEL", "OPENAI_API_KEY", "OPENAI_BASE_URL", "MIDSCENE_OPENAI_INIT_CONFIG_JSON", "MIDSCENE_OPENAI_HTTP_PROXY", "MIDSCENE_OPENAI_SOCKS_PROXY", "MIDSCENE_INSIGHT_MODEL_NAME", "MIDSCENE_INSIGHT_MODEL_SOCKS_PROXY", "MIDSCENE_INSIGHT_MODEL_HTTP_PROXY", "MIDSCENE_INSIGHT_MODEL_BASE_URL", "MIDSCENE_INSIGHT_MODEL_API_KEY", "MIDSCENE_INSIGHT_MODEL_INIT_CONFIG_JSON", "MIDSCENE_INSIGHT_MODEL_EXTRA_BODY_JSON", "MIDSCENE_INSIGHT_MODEL_TIMEOUT", "MIDSCENE_INSIGHT_MODEL_TEMPERATURE", "MIDSCENE_INSIGHT_MODEL_RETRY_COUNT", "MIDSCENE_INSIGHT_MODEL_RETRY_INTERVAL", "MIDSCENE_INSIGHT_MODEL_FAMILY", "MIDSCENE_INSIGHT_MODEL_REASONING_EFFORT", "MIDSCENE_INSIGHT_MODEL_REASONING_ENABLED", "MIDSCENE_INSIGHT_MODEL_REASONING_BUDGET", "MIDSCENE_PLANNING_MODEL_NAME", "MIDSCENE_PLANNING_MODEL_SOCKS_PROXY", "MIDSCENE_PLANNING_MODEL_HTTP_PROXY", "MIDSCENE_PLANNING_MODEL_BASE_URL", "MIDSCENE_PLANNING_MODEL_API_KEY", "MIDSCENE_PLANNING_MODEL_INIT_CONFIG_JSON", "MIDSCENE_PLANNING_MODEL_EXTRA_BODY_JSON", "MIDSCENE_PLANNING_MODEL_TIMEOUT", "MIDSCENE_PLANNING_MODEL_TEMPERATURE", "MIDSCENE_PLANNING_MODEL_RETRY_COUNT", "MIDSCENE_PLANNING_MODEL_RETRY_INTERVAL", "MIDSCENE_PLANNING_MODEL_FAMILY", "MIDSCENE_PLANNING_MODEL_REASONING_EFFORT", "MIDSCENE_PLANNING_MODEL_REASONING_ENABLED", "MIDSCENE_PLANNING_MODEL_REASONING_BUDGET", "MIDSCENE_MODEL_FAMILY"];
 export type TEnvKeys = (typeof ALL_ENV_KEYS)[number];
 export type TGlobalConfig = Record<TEnvKeys, string | undefined>;
 /**

package/dist/types/img/index.d.ts

@@ -1,3 +1,3 @@
 export { imageInfoOfBase64, isValidPNGImageBuffer, isValidJPEGImageBuffer, isValidImageBuffer, validateScreenshotBuffer, type ValidateScreenshotBufferOptions, } from './info';
-export { resizeAndConvertImgBuffer, resizeImgBase64, zoomForGPT4o, saveBase64Image, paddingToMatchBlockByBase64, cropByRect, scaleImage, localImg2Base64, httpImg2Base64, preProcessImageUrl, parseBase64, createImgBase64ByFormat, inferBase64ImageFormat, normalizeBase64Image, } from './transform';
+export { resizeAndConvertImgBuffer, resizeImgBase64, zoomForGPT4o, saveBase64Image, paddingToMatchBlockByBase64, cropByRect, scaleImage, localImg2Base64, httpImg2Base64, preProcessImageUrl, parseBase64, createImgBase64ByFormat, inferBase64ImageFormat, normalizeBase64Image, normalizeScreenshotBase64, type NormalizeScreenshotBase64Options, } from './transform';
 export { processImageElementInfo, compositeElementInfoImg, compositePointMarkerImg, annotateRects, } from './box-select';

package/dist/types/img/transform.d.ts

@@ -29,6 +29,10 @@ export declare function resizeAndConvertImgBuffer(inputFormat: string, inputData
 export declare const normalizeBase64Body: (body: string) => string;
 export declare const inferBase64ImageFormat: (base64Body: string) => "jpeg" | "png";
 export declare const createImgBase64ByFormat: (format: string, body: string) => string;
+export interface NormalizeScreenshotBase64Options {
+    label?: string;
+}
+export declare const normalizeScreenshotBase64: (base64: string, options?: NormalizeScreenshotBase64Options) => string;
 export declare const normalizeBase64Image: (base64: string) => string;
 export declare function resizeImgBase64(inputBase64: string, newSize: {
     width: number;

package/dist/types/key-alias-utils.d.ts

@@ -1,7 +1,7 @@
 /**
- * Internal-only helpers for CLI/MCP argument key aliasing.
+ * Internal-only helpers for CLI argument key aliasing.
  * Not re-exported from the package entry point — keep consumers within
- * `cli/` and `mcp/`.
+ * `cli/`.
  */
 export declare function kebabToCamel(str: string): string;
 export declare function camelToKebab(str: string): string;

package/dist/types/utils.d.ts

@@ -11,7 +11,6 @@ export declare function generateHashId(rect: any, content?: string): string;
  * @throws Error with the provided message if the condition is false
  */
 export declare function assert(condition: any, message?: string): asserts condition;
-export declare function setIsMcp(value: boolean): void;
 export declare function logMsg(...message: Parameters<typeof console.log>): void;
 export declare function repeat(times: number, fn: (index: number) => Promise<void>): Promise<void>;
 export declare const escapeScriptTag: (html: string) => string;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@midscene/shared",
-  "version": "1.9.7",
+  "version": "1.9.8-beta-20260618014851.0",
   "repository": "https://github.com/web-infra-dev/midscene",
   "homepage": "https://midscenejs.com/",
   "types": "./dist/types/index.d.ts",
@@ -57,16 +57,26 @@
       "import": "./dist/es/common.mjs",
       "require": "./dist/lib/common.js"
     },
-    "./mcp": {
-      "types": "./dist/types/mcp/index.d.ts",
-      "import": "./dist/es/mcp/index.mjs",
-      "require": "./dist/lib/mcp/index.js"
+    "./agent-tools": {
+      "types": "./dist/types/agent-tools/index.d.ts",
+      "import": "./dist/es/agent-tools/index.mjs",
+      "require": "./dist/lib/agent-tools/index.js"
+    },
+    "./agent-tools/*": {
+      "types": "./dist/types/agent-tools/*.d.ts",
+      "import": "./dist/es/agent-tools/*.mjs",
+      "require": "./dist/lib/agent-tools/*.js"
     },
     "./cli": {
       "types": "./dist/types/cli/index.d.ts",
       "import": "./dist/es/cli/index.mjs",
       "require": "./dist/lib/cli/index.js"
     },
+    "./recorder": {
+      "types": "./dist/types/recorder.d.ts",
+      "import": "./dist/es/recorder.mjs",
+      "require": "./dist/lib/recorder.js"
+    },
     "./logger": {
       "types": "./dist/types/logger.d.ts",
       "import": "./dist/es/logger.mjs",
@@ -84,11 +94,9 @@
     "README.md"
   ],
   "dependencies": {
-    "@modelcontextprotocol/sdk": "1.10.2",
     "@silvia-odwyer/photon": "0.3.3",
     "@silvia-odwyer/photon-node": "0.3.3",
     "debug": "4.4.0",
-    "express": "^4.21.2",
     "js-sha256": "0.11.0",
     "sharp": "^0.34.3",
     "dotenv": "^16.4.5",
@@ -98,7 +106,6 @@
   "devDependencies": {
     "@rslib/core": "^0.18.3",
     "@types/debug": "4.1.12",
-    "@types/express": "^4.17.21",
     "@types/node": "^18.0.0",
     "@ui-tars/shared": "1.2.0",
     "openai": "6.3.0",