@mastra/agent-browser 0.2.1-alpha.1 → 0.2.2-alpha.0

package/dist/index.d.cts

@@ -170,6 +170,13 @@ declare const dragInputSchema: z.ZodObject<{
     targetSelector: z.ZodOptional<z.ZodString>;
 }, z.core.$strip>;
 type DragInput = z.output<typeof dragInputSchema>;
+/**
+ * browser_screenshot - Capture a screenshot of the current page
+ */
+declare const screenshotInputSchema: z.ZodObject<{
+    fullPage: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>;
+type ScreenshotInput = z.output<typeof screenshotInputSchema>;
 /**
  * browser_evaluate - Execute JavaScript in the browser
  */
@@ -277,12 +284,38 @@ declare const browserSchemas: {
         sourceSelector: z.ZodOptional<z.ZodString>;
         targetSelector: z.ZodOptional<z.ZodString>;
     }, z.core.$strip>;
+    readonly screenshot: z.ZodObject<{
+        fullPage: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>;
     readonly evaluate: z.ZodObject<{
         script: z.ZodString;
         arg: z.ZodOptional<z.ZodUnknown>;
     }, z.core.$strip>;
 };
 
+/**
+ * Browser Tool Constants
+ */
+declare const BROWSER_TOOLS: {
+    readonly GOTO: "browser_goto";
+    readonly SNAPSHOT: "browser_snapshot";
+    readonly CLICK: "browser_click";
+    readonly TYPE: "browser_type";
+    readonly PRESS: "browser_press";
+    readonly SELECT: "browser_select";
+    readonly SCROLL: "browser_scroll";
+    readonly CLOSE: "browser_close";
+    readonly HOVER: "browser_hover";
+    readonly BACK: "browser_back";
+    readonly DIALOG: "browser_dialog";
+    readonly WAIT: "browser_wait";
+    readonly TABS: "browser_tabs";
+    readonly DRAG: "browser_drag";
+    readonly SCREENSHOT: "browser_screenshot";
+    readonly EVALUATE: "browser_evaluate";
+};
+type BrowserToolName = (typeof BROWSER_TOOLS)[keyof typeof BROWSER_TOOLS];
+
 /**
  * AgentBrowser-specific configuration extensions.
  */
@@ -300,6 +333,17 @@ interface AgentBrowserConfigExtensions {
      * ```
      */
     storageState?: string;
+    /**
+     * Tool names to exclude from the browser toolset.
+     * Use this to disable specific tools, e.g. `['browser_screenshot']`
+     * to skip the screenshot tool for models that don't support vision.
+     *
+     * @example
+     * ```ts
+     * new AgentBrowser({ excludeTools: ['browser_screenshot'] })
+     * ```
+     */
+    excludeTools?: BrowserToolName[];
 }
 /**
  * Configuration options for AgentBrowser.
@@ -387,6 +431,7 @@ declare class AgentBrowser extends MastraBrowser {
     private pidLookups;
     /** Thread manager - narrowed type from base class */
     protected threadManager: AgentBrowserThreadManager;
+    private browserConfig;
     constructor(config?: BrowserConfig);
     /**
      * Ensure browser is ready and thread session exists.
@@ -416,7 +461,7 @@ declare class AgentBrowser extends MastraBrowser {
     protected checkBrowserAlive(): Promise<boolean>;
     /**
      * Get the browser tools for this provider.
-     * Returns 17 flat tools for browser automation.
+     * Returns 16 flat tools for browser automation.
      */
     getTools(): Record<string, Tool<any, any>>;
     /**
@@ -497,6 +542,11 @@ declare class AgentBrowser extends MastraBrowser {
         scroll: string;
         hint?: string;
     } | BrowserToolError>;
+    screenshot(input: ScreenshotInput, threadId?: string): Promise<{
+        base64: string;
+        url: string;
+        title: string;
+    } | BrowserToolError>;
     click(input: ClickInput, threadId?: string): Promise<{
         success: true;
         url: string;
@@ -589,28 +639,6 @@ declare class AgentBrowser extends MastraBrowser {
  */
 declare function getBrowserPid(manager: BrowserManager): Promise<number | undefined>;
 
-/**
- * Browser Tool Constants
- */
-declare const BROWSER_TOOLS: {
-    readonly GOTO: "browser_goto";
-    readonly SNAPSHOT: "browser_snapshot";
-    readonly CLICK: "browser_click";
-    readonly TYPE: "browser_type";
-    readonly PRESS: "browser_press";
-    readonly SELECT: "browser_select";
-    readonly SCROLL: "browser_scroll";
-    readonly CLOSE: "browser_close";
-    readonly HOVER: "browser_hover";
-    readonly BACK: "browser_back";
-    readonly DIALOG: "browser_dialog";
-    readonly WAIT: "browser_wait";
-    readonly TABS: "browser_tabs";
-    readonly DRAG: "browser_drag";
-    readonly EVALUATE: "browser_evaluate";
-};
-type BrowserToolName = (typeof BROWSER_TOOLS)[keyof typeof BROWSER_TOOLS];
-
 /**
  * AgentBrowser Tools
  *

package/dist/index.d.ts

@@ -170,6 +170,13 @@ declare const dragInputSchema: z.ZodObject<{
     targetSelector: z.ZodOptional<z.ZodString>;
 }, z.core.$strip>;
 type DragInput = z.output<typeof dragInputSchema>;
+/**
+ * browser_screenshot - Capture a screenshot of the current page
+ */
+declare const screenshotInputSchema: z.ZodObject<{
+    fullPage: z.ZodOptional<z.ZodBoolean>;
+}, z.core.$strip>;
+type ScreenshotInput = z.output<typeof screenshotInputSchema>;
 /**
  * browser_evaluate - Execute JavaScript in the browser
  */
@@ -277,12 +284,38 @@ declare const browserSchemas: {
         sourceSelector: z.ZodOptional<z.ZodString>;
         targetSelector: z.ZodOptional<z.ZodString>;
     }, z.core.$strip>;
+    readonly screenshot: z.ZodObject<{
+        fullPage: z.ZodOptional<z.ZodBoolean>;
+    }, z.core.$strip>;
     readonly evaluate: z.ZodObject<{
         script: z.ZodString;
         arg: z.ZodOptional<z.ZodUnknown>;
     }, z.core.$strip>;
 };
 
+/**
+ * Browser Tool Constants
+ */
+declare const BROWSER_TOOLS: {
+    readonly GOTO: "browser_goto";
+    readonly SNAPSHOT: "browser_snapshot";
+    readonly CLICK: "browser_click";
+    readonly TYPE: "browser_type";
+    readonly PRESS: "browser_press";
+    readonly SELECT: "browser_select";
+    readonly SCROLL: "browser_scroll";
+    readonly CLOSE: "browser_close";
+    readonly HOVER: "browser_hover";
+    readonly BACK: "browser_back";
+    readonly DIALOG: "browser_dialog";
+    readonly WAIT: "browser_wait";
+    readonly TABS: "browser_tabs";
+    readonly DRAG: "browser_drag";
+    readonly SCREENSHOT: "browser_screenshot";
+    readonly EVALUATE: "browser_evaluate";
+};
+type BrowserToolName = (typeof BROWSER_TOOLS)[keyof typeof BROWSER_TOOLS];
+
 /**
  * AgentBrowser-specific configuration extensions.
  */
@@ -300,6 +333,17 @@ interface AgentBrowserConfigExtensions {
      * ```
      */
     storageState?: string;
+    /**
+     * Tool names to exclude from the browser toolset.
+     * Use this to disable specific tools, e.g. `['browser_screenshot']`
+     * to skip the screenshot tool for models that don't support vision.
+     *
+     * @example
+     * ```ts
+     * new AgentBrowser({ excludeTools: ['browser_screenshot'] })
+     * ```
+     */
+    excludeTools?: BrowserToolName[];
 }
 /**
  * Configuration options for AgentBrowser.
@@ -387,6 +431,7 @@ declare class AgentBrowser extends MastraBrowser {
     private pidLookups;
     /** Thread manager - narrowed type from base class */
     protected threadManager: AgentBrowserThreadManager;
+    private browserConfig;
     constructor(config?: BrowserConfig);
     /**
      * Ensure browser is ready and thread session exists.
@@ -416,7 +461,7 @@ declare class AgentBrowser extends MastraBrowser {
     protected checkBrowserAlive(): Promise<boolean>;
     /**
      * Get the browser tools for this provider.
-     * Returns 17 flat tools for browser automation.
+     * Returns 16 flat tools for browser automation.
      */
     getTools(): Record<string, Tool<any, any>>;
     /**
@@ -497,6 +542,11 @@ declare class AgentBrowser extends MastraBrowser {
         scroll: string;
         hint?: string;
     } | BrowserToolError>;
+    screenshot(input: ScreenshotInput, threadId?: string): Promise<{
+        base64: string;
+        url: string;
+        title: string;
+    } | BrowserToolError>;
     click(input: ClickInput, threadId?: string): Promise<{
         success: true;
         url: string;
@@ -589,28 +639,6 @@ declare class AgentBrowser extends MastraBrowser {
  */
 declare function getBrowserPid(manager: BrowserManager): Promise<number | undefined>;
 
-/**
- * Browser Tool Constants
- */
-declare const BROWSER_TOOLS: {
-    readonly GOTO: "browser_goto";
-    readonly SNAPSHOT: "browser_snapshot";
-    readonly CLICK: "browser_click";
-    readonly TYPE: "browser_type";
-    readonly PRESS: "browser_press";
-    readonly SELECT: "browser_select";
-    readonly SCROLL: "browser_scroll";
-    readonly CLOSE: "browser_close";
-    readonly HOVER: "browser_hover";
-    readonly BACK: "browser_back";
-    readonly DIALOG: "browser_dialog";
-    readonly WAIT: "browser_wait";
-    readonly TABS: "browser_tabs";
-    readonly DRAG: "browser_drag";
-    readonly EVALUATE: "browser_evaluate";
-};
-type BrowserToolName = (typeof BROWSER_TOOLS)[keyof typeof BROWSER_TOOLS];
-
 /**
  * AgentBrowser Tools
  *

package/dist/index.js

@@ -218,6 +218,9 @@ var dragInputSchema = z.object({
     });
   }
 });
+var screenshotInputSchema = z.object({
+  fullPage: z.boolean().optional().describe("Capture the full scrollable page instead of just the viewport (default: false)")
+});
 var evaluateInputSchema = z.object({
   script: z.string().describe(
     "JavaScript expression to evaluate in the browser and return the result. Do not use `return` \u2014 write a bare expression like `document.title` or `1 + 1`. For async code, wrap in an async IIFE: `(async () => { ... })()`."
@@ -241,6 +244,8 @@ var browserSchemas = {
   wait: waitInputSchema,
   tabs: tabsInputSchema,
   drag: dragInputSchema,
+  // Utility
+  screenshot: screenshotInputSchema,
   // Escape hatch
   evaluate: evaluateInputSchema
 };
@@ -263,6 +268,8 @@ var BROWSER_TOOLS = {
   WAIT: "browser_wait",
   TABS: "browser_tabs",
   DRAG: "browser_drag",
+  // Utility
+  SCREENSHOT: "browser_screenshot",
   // Escape hatch
   EVALUATE: "browser_evaluate"
 };
@@ -391,6 +398,38 @@ function createPressTool(browser) {
     }
   });
 }
+function createScreenshotTool(browser) {
+  return createTool({
+    id: BROWSER_TOOLS.SCREENSHOT,
+    description: "Capture a screenshot of the current viewport as a visible PNG (set fullPage: true for full-page capture). Use snapshot when you only need text or interactive elements \u2014 screenshots are expensive. Use this when you need to visually inspect the page, e.g. evaluating images, product photos, layout, design, or colors.",
+    inputSchema: screenshotInputSchema,
+    execute: async (input, { agent }) => {
+      const threadId = agent?.threadId;
+      browser.setCurrentThread(threadId);
+      await browser.ensureReady();
+      return await browser.screenshot(input, threadId);
+    },
+    toModelOutput(output) {
+      const result = output;
+      if (typeof result.base64 !== "string") {
+        return {
+          type: "content",
+          value: [{ type: "text", text: result.message ?? "Failed to capture screenshot." }]
+        };
+      }
+      return {
+        type: "content",
+        value: [
+          {
+            type: "media",
+            mediaType: "image/png",
+            data: result.base64
+          }
+        ]
+      };
+    }
+  });
+}
 function createScrollTool(browser) {
   return createTool({
     id: BROWSER_TOOLS.SCROLL,
@@ -482,6 +521,8 @@ function createAgentBrowserTools(browser) {
     [BROWSER_TOOLS.SELECT]: createSelectTool(browser),
     [BROWSER_TOOLS.SCROLL]: createScrollTool(browser),
     [BROWSER_TOOLS.CLOSE]: createCloseTool(browser),
+    // Utility
+    [BROWSER_TOOLS.SCREENSHOT]: createScreenshotTool(browser),
     // Extended
     [BROWSER_TOOLS.HOVER]: createHoverTool(browser),
     [BROWSER_TOOLS.BACK]: createBackTool(browser),
@@ -524,8 +565,10 @@ var AgentBrowser = class extends MastraBrowser {
   defaultTimeout = 3e4;
   /** Pending PID lookups — awaited in disconnect handlers to avoid racing. */
   pidLookups = /* @__PURE__ */ new Set();
+  browserConfig;
   constructor(config = {}) {
     super(config);
+    this.browserConfig = config;
     this.id = `agent-browser-${Date.now()}`;
     if (config.timeout) {
       this.defaultTimeout = config.timeout;
@@ -688,10 +731,17 @@ var AgentBrowser = class extends MastraBrowser {
   // ---------------------------------------------------------------------------
   /**
    * Get the browser tools for this provider.
-   * Returns 17 flat tools for browser automation.
+   * Returns 16 flat tools for browser automation.
    */
   getTools() {
-    return createAgentBrowserTools(this);
+    const tools = createAgentBrowserTools(this);
+    const exclude = this.browserConfig.excludeTools;
+    if (exclude?.length) {
+      for (const name of exclude) {
+        delete tools[name];
+      }
+    }
+    return tools;
   }
   // ---------------------------------------------------------------------------
   // Helpers
@@ -1004,6 +1054,26 @@ var AgentBrowser = class extends MastraBrowser {
     }
   }
   // ---------------------------------------------------------------------------
+  // browser_screenshot - Capture a screenshot of the current page
+  // ---------------------------------------------------------------------------
+  async screenshot(input, threadId) {
+    try {
+      const page = await this.getPage(threadId);
+      const buffer = await page.screenshot({
+        fullPage: input.fullPage ?? false,
+        type: "png"
+      });
+      const base64 = Buffer.from(buffer).toString("base64");
+      return {
+        base64,
+        url: page.url(),
+        title: await page.title()
+      };
+    } catch (error) {
+      return this.createErrorFromException(error, "Screenshot");
+    }
+  }
+  // ---------------------------------------------------------------------------
   // 3. browser_click - Click on element
   // ---------------------------------------------------------------------------
   async click(input, threadId) {