ornold-mcp 1.2.2 → 1.3.2

package/dist/cli.js

@@ -2056,6 +2056,7 @@ function getOptionalString(argNames, envNames) {
 }
 var TOKEN = getArg("token") || process.env.ORNOLD_TOKEN || "";
 var SERVER_URL = getArg("server") || process.env.ORNOLD_SERVER || "wss://mcp.ornold.com/bridge";
+var BROWSER_MODE = getArg("mode") || process.env.ORNOLD_MODE || "dom";
 var LINKEN_PORT = getOptionalPort("linken-port", "LINKEN_PORT");
 var WADEX_PORT = getOptionalPort("wadex-port", "WADEX_PORT");
 var DOLPHIN_PORT = getOptionalPort("dolphin-port", "DOLPHIN_PORT");
@@ -2432,8 +2433,19 @@ ${result.snapshot || ""}`;
       }
       case "cdp_click_normalized_box": {
         const exec = await getExecutor();
-        const r = await exec.parallelClickNormalizedBox(params.box, params.browserIds);
-        return { content: formatResult(r), isError: false };
+        const box = params.box;
+        if (!box || box.length !== 4) return { content: "Invalid box coordinates", isError: true };
+        const browsers = params.browserIds ? void 0 : await exec.listBrowsers();
+        const targetId = params.browserIds?.[0] || browsers?.[0]?.id;
+        if (!targetId) return { content: "No browser connected", isError: true };
+        const page = exec.getPage(targetId);
+        if (!page) return { content: "Browser page not found", isError: true };
+        const vpRaw = await page.evaluate("JSON.stringify({ w: window.innerWidth, h: window.innerHeight })");
+        const vp = JSON.parse(vpRaw);
+        const centerX = (box[0] + box[2]) / 2 * vp.w;
+        const centerY = (box[1] + box[3]) / 2 * vp.h;
+        await page.mouse.click(centerX, centerY);
+        return { content: JSON.stringify({ clicked: true, x: Math.round(centerX), y: Math.round(centerY), viewport: vp }), isError: false };
       }
       case "cdp_setup_downloads": {
         const exec = await getExecutor();
@@ -2865,6 +2877,16 @@ var browserTargetArgs = {
   _scope: browserScopeArg
 };
 function createServer() {
+  const modeInstructions = BROWSER_MODE === "vision" ? `## Interaction mode: VISION
+- Use browser_parallel_vision_analyze_grouped to analyze page content via AI vision.
+- Use browser_parallel_click_normalized_box to click elements by coordinates.
+- Use browser_parallel_screenshot to see the page.
+- Do NOT use browser_parallel_snapshot (DOM mode is disabled).` : BROWSER_MODE === "both" ? `## Interaction modes: DOM + VISION
+- Prefer DOM mode (browser_parallel_snapshot \u2192 ref-based clicks) \u2014 it's faster and free.
+- Use Vision mode (browser_parallel_vision_analyze_grouped) only when DOM doesn't work (canvas, complex iframes, dynamic content).` : `## Interaction mode: DOM
+- Use browser_parallel_snapshot to get page content with [ref=N] markers.
+- Use ref parameter for clicking/filling elements from snapshot.
+- Always call snapshot BEFORE clicking/filling \u2014 refs go stale after navigation.`;
   const server = new McpServer({
     name: "ornold-browser",
     version: CLIENT_VERSION,
@@ -2876,9 +2898,9 @@ ${LINKEN_PORT !== void 0 ? "- Linken Sphere (linken_* tools)" : ""}${WADEX_PORT
 ## Core workflow
 1. Start a browser profile: linken_start_instances / dolphin_start_profile
 2. Browser auto-connects via CDP. Use browser_list to see connected browsers.
-3. Use browser_parallel_* tools to interact (navigate, click, fill, snapshot, etc.)
-4. Always call browser_parallel_snapshot BEFORE clicking/filling \u2014 it returns [ref=N] markers.
-5. Use ref parameter (not selector) for clicking/filling elements from snapshot.
+3. Use browser_parallel_* tools to interact with pages.
+
+${modeInstructions}
 
 ## Anti-detection rules
 - Navigate via Google search, not direct URLs (except same-domain links).
@@ -2886,12 +2908,9 @@ ${LINKEN_PORT !== void 0 ? "- Linken Sphere (linken_* tools)" : ""}${WADEX_PORT
 - Use browser_parallel_type (not fill) for short human-like inputs.
 - Never modify fingerprints via JavaScript or navigate to chrome:// URLs.
 
-## Flow execution (AI-powered automation)
-Flows delegate multi-step tasks to a cheaper model on the server.
 ## Captcha solving
 - browser_detect_captcha \u2192 browser_solve_captcha (reCAPTCHA/hCaptcha via 2captcha)
-- browser_detect_press_hold \u2192 browser_solve_press_hold (PerimeterX)
-- browser_captcha_balance to check remaining balance`
+- browser_detect_press_hold \u2192 browser_solve_press_hold (PerimeterX)`
   });
   server.tool("browser_list", "List connected browsers", {}, () => toolHandler("browser_list", {}));
   server.tool("browser_status", "Check browser sync and responsiveness", {
@@ -2905,11 +2924,29 @@ Flows delegate multi-step tasks to a cheaper model on the server.
     browserIds: browserIdsArg,
     _scope: browserScopeArg
   }, (args2) => toolHandler("browser_parallel_tabs", args2));
-  server.tool("browser_parallel_snapshot", "Get page snapshot with [ref=N] markers", {
-    compact: z.boolean().optional(),
-    browserIds: browserIdsArg,
-    _scope: browserScopeArg
-  }, (args2) => toolHandler("browser_parallel_snapshot", args2));
+  if (BROWSER_MODE === "dom" || BROWSER_MODE === "both") {
+    server.tool("browser_parallel_snapshot", "Get page snapshot with [ref=N] markers", {
+      compact: z.boolean().optional(),
+      browserIds: browserIdsArg,
+      _scope: browserScopeArg
+    }, (args2) => toolHandler("browser_parallel_snapshot", args2));
+  }
+  if (BROWSER_MODE === "vision" || BROWSER_MODE === "both") {
+    server.tool("browser_parallel_screenshot", "Take a screenshot of the page", {
+      browserIds: browserIdsArg,
+      _scope: browserScopeArg
+    }, (args2) => toolHandler("browser_parallel_screenshot", args2));
+    server.tool("browser_parallel_vision_analyze_grouped", "Analyze grouped browser screenshots with OmniParser via Ornold server", {
+      similarityThreshold: z.number().optional(),
+      browserIds: browserIdsArg,
+      _scope: browserScopeArg
+    }, (args2) => toolHandler("browser_parallel_vision_analyze_grouped", args2));
+    server.tool("browser_parallel_click_normalized_box", "Click normalized viewport box center", {
+      box: z.tuple([z.number(), z.number(), z.number(), z.number()]),
+      browserIds: browserIdsArg,
+      _scope: browserScopeArg
+    }, (args2) => toolHandler("browser_parallel_click_normalized_box", args2));
+  }
   server.tool("browser_parallel_navigate", "Navigate to URL", {
     url: z.string(),
     browserIds: browserIdsArg,
@@ -2922,52 +2959,57 @@ Flows delegate multi-step tasks to a cheaper model on the server.
     })),
     _scope: browserScopeArg
   }, (args2) => toolHandler("browser_parallel_navigate_multi", args2));
-  server.tool("browser_parallel_click", "Click element", {
-    ...browserTargetArgs
-  }, (args2) => toolHandler("browser_parallel_click", args2));
-  server.tool("browser_parallel_type", "Type text into element", {
-    ...browserTargetArgs,
-    text: z.string()
-  }, (args2) => toolHandler("browser_parallel_type", args2));
-  server.tool("browser_parallel_fill", "Fill input (clear + type)", {
-    ...browserTargetArgs,
-    text: z.string().optional(),
-    texts: z.record(z.string()).optional()
-  }, (args2) => toolHandler("browser_parallel_fill", args2));
-  server.tool("browser_parallel_fill_multi", "Fill input with per-browser values", {
-    ...browserTargetArgs,
-    texts: z.record(z.string())
-  }, (args2) => toolHandler("browser_parallel_fill_multi", args2));
-  server.tool("browser_parallel_fill_form", "Fill multiple form fields sequentially", {
-    fields: z.array(z.object({
-      element: z.string().optional(),
-      ref: browserRefArg,
-      selector: z.string().optional(),
-      value: z.string(),
-      type: z.enum(["textbox", "checkbox", "radio", "combobox"]).optional()
-    })),
-    browserIds: browserIdsArg,
-    _scope: browserScopeArg
-  }, (args2) => toolHandler("browser_parallel_fill_form", args2));
-  server.tool("browser_parallel_drag", "Drag from one element to another", {
-    startElement: z.string().optional(),
-    startRef: browserRefArg,
-    startSelector: z.string().optional(),
-    endElement: z.string().optional(),
-    endRef: browserRefArg,
-    endSelector: z.string().optional(),
-    browserIds: browserIdsArg,
-    _scope: browserScopeArg
-  }, (args2) => toolHandler("browser_parallel_drag", args2));
+  if (BROWSER_MODE === "dom" || BROWSER_MODE === "both") {
+    server.tool("browser_parallel_click", "Click element", {
+      ...browserTargetArgs
+    }, (args2) => toolHandler("browser_parallel_click", args2));
+    server.tool("browser_parallel_type", "Type text into element", {
+      ...browserTargetArgs,
+      text: z.string()
+    }, (args2) => toolHandler("browser_parallel_type", args2));
+    server.tool("browser_parallel_fill", "Fill input (clear + type)", {
+      ...browserTargetArgs,
+      text: z.string().optional(),
+      texts: z.record(z.string()).optional()
+    }, (args2) => toolHandler("browser_parallel_fill", args2));
+    server.tool("browser_parallel_fill_multi", "Fill input with per-browser values", {
+      ...browserTargetArgs,
+      texts: z.record(z.string())
+    }, (args2) => toolHandler("browser_parallel_fill_multi", args2));
+    server.tool("browser_parallel_fill_form", "Fill multiple form fields sequentially", {
+      fields: z.array(z.object({
+        element: z.string().optional(),
+        ref: browserRefArg,
+        selector: z.string().optional(),
+        value: z.string(),
+        type: z.enum(["textbox", "checkbox", "radio", "combobox"]).optional()
+      })),
+      browserIds: browserIdsArg,
+      _scope: browserScopeArg
+    }, (args2) => toolHandler("browser_parallel_fill_form", args2));
+    server.tool("browser_parallel_drag", "Drag from one element to another", {
+      startElement: z.string().optional(),
+      startRef: browserRefArg,
+      startSelector: z.string().optional(),
+      endElement: z.string().optional(),
+      endRef: browserRefArg,
+      endSelector: z.string().optional(),
+      browserIds: browserIdsArg,
+      _scope: browserScopeArg
+    }, (args2) => toolHandler("browser_parallel_drag", args2));
+    server.tool("browser_parallel_select_option", "Select dropdown option", {
+      ...browserTargetArgs,
+      values: z.array(z.string())
+    }, (args2) => toolHandler("browser_parallel_select_option", args2));
+    server.tool("browser_parallel_hover", "Hover over element", {
+      ...browserTargetArgs
+    }, (args2) => toolHandler("browser_parallel_hover", args2));
+  }
   server.tool("browser_parallel_press_key", "Press keyboard key", {
     key: z.string(),
     browserIds: browserIdsArg,
     _scope: browserScopeArg
   }, (args2) => toolHandler("browser_parallel_press_key", args2));
-  server.tool("browser_parallel_select_option", "Select dropdown option", {
-    ...browserTargetArgs,
-    values: z.array(z.string())
-  }, (args2) => toolHandler("browser_parallel_select_option", args2));
   server.tool("browser_parallel_wait_for", "Wait for condition", {
     time: z.number().optional(),
     text: z.string().optional(),
@@ -2977,11 +3019,6 @@ Flows delegate multi-step tasks to a cheaper model on the server.
     browserIds: browserIdsArg,
     _scope: browserScopeArg
   }, (args2) => toolHandler("browser_parallel_wait_for", args2));
-  server.tool("browser_parallel_screenshot", "Take screenshot", {
-    fullPage: z.boolean().optional(),
-    browserIds: browserIdsArg,
-    _scope: browserScopeArg
-  }, (args2) => toolHandler("browser_parallel_screenshot", args2));
   server.tool("browser_parallel_evaluate", "Run JavaScript in page", {
     script: z.string(),
     browserIds: browserIdsArg,
@@ -2997,9 +3034,6 @@ Flows delegate multi-step tasks to a cheaper model on the server.
     variables: z.record(z.record(z.string())),
     _scope: browserScopeArg
   }, (args2) => toolHandler("browser_parallel_run_code_with_vars", args2));
-  server.tool("browser_parallel_hover", "Hover over element", {
-    ...browserTargetArgs
-  }, (args2) => toolHandler("browser_parallel_hover", args2));
   server.tool("browser_parallel_go_back", "Go back", {
     browserIds: browserIdsArg,
     _scope: browserScopeArg
@@ -3027,16 +3061,6 @@ Flows delegate multi-step tasks to a cheaper model on the server.
     browserIds: browserIdsArg,
     _scope: browserScopeArg
   }, (args2) => toolHandler("browser_parallel_network_requests", args2));
-  server.tool("browser_parallel_vision_analyze_grouped", "Analyze grouped browser screenshots with OmniParser via Ornold server", {
-    similarityThreshold: z.number().optional(),
-    browserIds: browserIdsArg,
-    _scope: browserScopeArg
-  }, (args2) => toolHandler("browser_parallel_vision_analyze_grouped", args2));
-  server.tool("browser_parallel_click_normalized_box", "Click normalized viewport box center", {
-    box: z.tuple([z.number(), z.number(), z.number(), z.number()]),
-    browserIds: browserIdsArg,
-    _scope: browserScopeArg
-  }, (args2) => toolHandler("browser_parallel_click_normalized_box", args2));
   server.tool("browser_setup_downloads", "Enable browser downloads to project files directory", {
     browserIds: browserIdsArg,
     _scope: browserScopeArg
@@ -3083,7 +3107,6 @@ Flows delegate multi-step tasks to a cheaper model on the server.
     autoSubmit: z.boolean().optional(),
     _scope: browserScopeArg
   }, (args2) => toolHandler("browser_solve_captcha", args2));
-  server.tool("browser_captcha_balance", "Check 2captcha balance", {}, () => toolHandler("browser_captcha_balance", {}));
   server.tool("captcha_detect", "Detect captcha on page", {
     browserIds: browserIdsArg
   }, (args2) => toolHandler("captcha_detect", args2));

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "ornold-mcp",
-  "version": "1.2.2",
+  "version": "1.3.2",
   "type": "module",
   "main": "./dist/cli.js",
   "files": [
@@ -19,7 +19,6 @@
     "client": "tsx client/index.ts"
   },
   "dependencies": {
-    "@anthropic-ai/sdk": "^0.88.0",
     "@modelcontextprotocol/sdk": "^1.12.0",
     "@supabase/supabase-js": "^2.103.0",
     "express": "^4.21.2",