diff --git a/src/DevtoolsUtils.ts b/src/DevtoolsUtils.ts
index 7a4abe9dc..c1f6efcb4 100644
--- a/src/DevtoolsUtils.ts
+++ b/src/DevtoolsUtils.ts
@@ -15,6 +15,61 @@ import type {
   Target as PuppeteerTarget,
 } from './third_party/index.js';
 
+export function extractUrlLikeFromDevToolsTitle(
+  title: string,
+): string | undefined {
+  const match = title.match(new RegExp(`DevTools - (.*)`));
+  return match?.[1] ?? undefined;
+}
+
+export function urlsEqual(url1: string, url2: string): boolean {
+  const normalizedUrl1 = normalizeUrl(url1);
+  const normalizedUrl2 = normalizeUrl(url2);
+  return normalizedUrl1 === normalizedUrl2;
+}
+
+/**
+ * For the sake of the MCP server, when we determine if two URLs are equal we
+ * remove some parts:
+ *
+ * 1. We do not care about the protocol.
+ * 2. We do not care about trailing slashes.
+ * 3. We do not care about "www".
+ * 4. We ignore the hash parts.
+ *
+ * For example, if the user types "record a trace on foo.com", we would want to
+ * match a tab in the connected Chrome instance that is showing "www.foo.com/"
+ */
+function normalizeUrl(url: string): string {
+  let result = url.trim();
+
+  // Remove protocols
+  if (result.startsWith('https://')) {
+    result = result.slice(8);
+  } else if (result.startsWith('http://')) {
+    result = result.slice(7);
+  }
+
+  // Remove 'www.'. This ensures that we find the right URL regardless of if the user adds `www` or not.
+  if (result.startsWith('www.')) {
+    result = result.slice(4);
+  }
+
+  // We use target URLs to locate DevTools but those often do
+  // no include hash.
+  const hashIdx = result.lastIndexOf('#');
+  if (hashIdx !== -1) {
+    result = result.slice(0, hashIdx);
+  }
+
+  // Remove trailing slash
+  if (result.endsWith('/')) {
+    result = result.slice(0, -1);
+  }
+
+  return result;
+}
+
 /**
  * A mock implementation of an issues manager that only implements the methods
  * that are actually used by the IssuesAggregator
diff --git a/src/McpContext.ts b/src/McpContext.ts
index 4647c6118..889103fab 100644
--- a/src/McpContext.ts
+++ b/src/McpContext.ts
@@ -8,7 +8,11 @@ import fs from 'node:fs/promises';
 import path from 'node:path';
 
 import type {TargetUniverse} from './DevtoolsUtils.js';
-import {UniverseManager} from './DevtoolsUtils.js';
+import {
+  extractUrlLikeFromDevToolsTitle,
+  UniverseManager,
+  urlsEqual,
+} from './DevtoolsUtils.js';
 import {McpPage} from './McpPage.js';
 import type {ListenerMap, UncaughtError} from './PageCollector.js';
 import {NetworkCollector, ConsoleCollector} from './PageCollector.js';
@@ -649,21 +653,37 @@ export class McpContext implements Context {
   async detectOpenDevToolsWindows() {
     this.logger('Detecting open DevTools windows');
     const {pages} = await this.#getAllPages();
-
-    await Promise.all(
-      pages.map(async page => {
-        const mcpPage = this.#mcpPages.get(page);
-        if (!mcpPage) {
-          return;
-        }
-
-        if (await page.hasDevTools()) {
-          mcpPage.devToolsPage = await page.openDevTools();
-        } else {
-          mcpPage.devToolsPage = undefined;
+    // Clear all devToolsPage references before re-detecting.
+    for (const mcpPage of this.#mcpPages.values()) {
+      mcpPage.devToolsPage = undefined;
+    }
+    for (const devToolsPage of pages) {
+      if (devToolsPage.url().startsWith('devtools://')) {
+        try {
+          this.logger('Calling getTargetInfo for ' + devToolsPage.url());
+          const data = await devToolsPage
+            // @ts-expect-error no types for _client().
+            ._client()
+            .send('Target.getTargetInfo');
+          const devtoolsPageTitle = data.targetInfo.title;
+          const urlLike = extractUrlLikeFromDevToolsTitle(devtoolsPageTitle);
+          if (!urlLike) {
+            continue;
+          }
+          // TODO: lookup without a loop.
+          for (const page of this.#pages) {
+            if (urlsEqual(page.url(), urlLike)) {
+              const mcpPage = this.#mcpPages.get(page);
+              if (mcpPage) {
+                mcpPage.devToolsPage = devToolsPage;
+              }
+            }
+          }
+        } catch (error) {
+          this.logger('Issue occurred while trying to find DevTools', error);
         }
-      }),
-    );
+      }
+    }
   }
 
   getExtensionServiceWorkers(): ExtensionServiceWorker[] {
diff --git a/tests/DevtoolsUtils.test.ts b/tests/DevtoolsUtils.test.ts
index 8f9aa3794..a5e43009c 100644
--- a/tests/DevtoolsUtils.test.ts
+++ b/tests/DevtoolsUtils.test.ts
@@ -9,7 +9,11 @@ import {afterEach, describe, it} from 'node:test';
 
 import sinon from 'sinon';
 
-import {UniverseManager} from '../src/DevtoolsUtils.js';
+import {
+  extractUrlLikeFromDevToolsTitle,
+  urlsEqual,
+  UniverseManager,
+} from '../src/DevtoolsUtils.js';
 import {DevTools} from '../src/third_party/index.js';
 import type {Browser, Target} from '../src/third_party/index.js';
 
@@ -20,6 +24,76 @@ import {
   withBrowser,
 } from './utils.js';
 
+describe('extractUrlFromDevToolsTitle', () => {
+  it('deals with no trailing /', () => {
+    assert.strictEqual(
+      extractUrlLikeFromDevToolsTitle('DevTools - example.com'),
+      'example.com',
+    );
+  });
+  it('deals with a trailing /', () => {
+    assert.strictEqual(
+      extractUrlLikeFromDevToolsTitle('DevTools - example.com/'),
+      'example.com/',
+    );
+  });
+  it('deals with www', () => {
+    assert.strictEqual(
+      extractUrlLikeFromDevToolsTitle('DevTools - www.example.com/'),
+      'www.example.com/',
+    );
+  });
+  it('deals with complex url', () => {
+    assert.strictEqual(
+      extractUrlLikeFromDevToolsTitle(
+        'DevTools - www.example.com/path.html?a=b#3',
+      ),
+      'www.example.com/path.html?a=b#3',
+    );
+  });
+});
+
+describe('urlsEqual', () => {
+  it('ignores trailing slashes', () => {
+    assert.strictEqual(
+      urlsEqual('https://google.com/', 'https://google.com'),
+      true,
+    );
+  });
+
+  it('ignores www', () => {
+    assert.strictEqual(
+      urlsEqual('https://google.com/', 'https://www.google.com'),
+      true,
+    );
+  });
+
+  it('ignores protocols', () => {
+    assert.strictEqual(
+      urlsEqual('https://google.com/', 'http://www.google.com'),
+      true,
+    );
+  });
+
+  it('does not ignore other subdomains', () => {
+    assert.strictEqual(
+      urlsEqual('https://google.com/', 'https://photos.google.com'),
+      false,
+    );
+  });
+
+  it('ignores hash', () => {
+    assert.strictEqual(
+      urlsEqual('https://google.com/#', 'http://www.google.com'),
+      true,
+    );
+    assert.strictEqual(
+      urlsEqual('https://google.com/#21', 'http://www.google.com#12'),
+      true,
+    );
+  });
+});
+
 describe('UniverseManager', () => {
   afterEach(() => {
     sinon.restore();