@everydaydevopsio/pilot 0.6.2 → 0.7.0

package/src/mcp/tools/browser.ts

@@ -1,26 +1,14 @@
 import { z } from 'zod';
 import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
-import type { Client } from 'chrome-remote-interface';
-import type { BrowserManager } from '../../browser.js';
-import type { BrowserContext } from '../server.js';
-import { executeScreenshot } from '../../commands/screenshot.js';
-import { executeNavigate } from '../../commands/navigate.js';
-import { executeClick } from '../../commands/click.js';
-import { executeType } from '../../commands/type.js';
-import { executeEvaluate } from '../../commands/evaluate.js';
-import { executeWait } from '../../commands/wait.js';
-import { executePageInfo } from '../../commands/page_info.js';
+import type { BrowserContext, DriverFactory } from '../server.js';
+import { DEFAULT_STARTUP_TIMEOUT_MS } from '../../browser/chrome-launcher.js';
 import {
-  executeListTabs,
-  executeNewTab,
-  executeCloseTab,
-  executeSwitchTab
-} from '../../commands/tabs.js';
-import {
-  VIEWPORT_PRESETS,
-  applyViewport,
-  type ViewportConfig
-} from '../../viewport.js';
+  listProfiles,
+  pilotDataDir,
+  stopPilotBrowsers
+} from '../../browser/sessions.js';
+import { VIEWPORT_PRESETS } from '../../viewport.js';
+import { requireDriver } from './driver-context.js';
 
 // Raw shapes for MCP SDK (expects ZodRawShape, not ZodObject)
 
@@ -185,9 +173,9 @@ const startShape = {
     .number()
     .int()
     .positive()
-    .default(30000)
+    .default(DEFAULT_STARTUP_TIMEOUT_MS)
     .describe(
-      'Timeout waiting for Chrome DevTools to become ready, in milliseconds.'
+      `Timeout waiting for Chrome DevTools to become ready, in milliseconds. Defaults to ${DEFAULT_STARTUP_TIMEOUT_MS}. Startup fails immediately if Chrome exits, so raising this only helps when Chrome is slow to start (for example a first run on a Linux server with software rendering).`
     )
 };
 
@@ -242,22 +230,10 @@ const viewportResizeShape = {
     )
 };
 
-function requireClient(context: BrowserContext): {
-  client: Client;
-  manager: BrowserManager;
-} {
-  const manager = context.manager;
-  if (!manager || !manager.isConnected()) {
-    throw new Error('Browser not started. Call browser_start first.');
-  }
-  const client = manager.getClient()!;
-  return { client, manager };
-}
-
 export function registerBrowserTools(
   server: McpServer,
   context: BrowserContext,
-  makeBrowserManager: () => BrowserManager
+  makeDriver: DriverFactory
 ): void {
   server.tool(
     'browser_start',
@@ -274,7 +250,7 @@ export function registerBrowserTools(
       responsive,
       startupTimeoutMs
     }) => {
-      if (context.manager?.isConnected()) {
+      if (context.driver?.isConnected()) {
         return {
           content: [
             {
@@ -285,14 +261,14 @@ export function registerBrowserTools(
         };
       }
 
-      if (context.manager) {
-        await context.manager.destroy();
-        context.manager = null;
+      if (context.driver) {
+        await context.driver.destroy();
+        context.driver = null;
       }
 
-      const manager = makeBrowserManager();
+      const driver = makeDriver();
       const { headless: effectiveHeadless, viewport: effectiveViewport } =
-        await manager.launch({
+        await driver.launch({
           headless,
           chromePath,
           profileName,
@@ -303,7 +279,7 @@ export function registerBrowserTools(
           responsive,
           startupTimeoutMs
         });
-      context.manager = manager;
+      context.driver = driver;
 
       return {
         content: [
@@ -321,14 +297,19 @@ export function registerBrowserTools(
     'Stop the Chrome browser and free all resources.',
     {},
     async () => {
-      if (!context.manager) {
+      if (!context.driver) {
         return {
-          content: [{ type: 'text' as const, text: 'Browser is not running.' }]
+          content: [
+            {
+              type: 'text' as const,
+              text: 'Browser is not running. This Pilot session is not tracking a launched or connected browser; a failed browser_start stops the Chrome it launched and reports the outcome in its error.'
+            }
+          ]
         };
       }
 
-      await context.manager.destroy();
-      context.manager = null;
+      await context.driver.destroy();
+      context.driver = null;
 
       return {
         content: [{ type: 'text' as const, text: 'Browser stopped.' }]
@@ -336,13 +317,70 @@ export function registerBrowserTools(
     }
   );
 
+  server.tool(
+    'browser_status',
+    'Show the browser tracked by this Pilot session and the lock state of every Pilot profile on this machine (including browsers started by other Pilot sessions). Read-only. Use it to diagnose "profile already in use" errors.',
+    {},
+    async () => {
+      const status = {
+        session: context.driver?.getSessionInfo() ?? {
+          mode: 'none',
+          connected: false
+        },
+        dataDir: pilotDataDir(),
+        profiles: listProfiles()
+      };
+      return {
+        content: [
+          { type: 'text' as const, text: JSON.stringify(status, null, 2) }
+        ]
+      };
+    }
+  );
+
+  server.tool(
+    'browser_stop_all',
+    'Stop every Pilot-managed Chrome on this machine: the browser in this session plus any Chrome holding a Pilot profile lock, including ones started by other Pilot sessions. Only processes whose command line shows Chrome using a Pilot profile directory are stopped; browsers on other hosts and unverifiable locks are skipped. Also removes stale locks. Ask the user before calling this, since it can close browsers other agents are using.',
+    {},
+    async () => {
+      const lines: string[] = [];
+      if (context.driver) {
+        const info = context.driver.getSessionInfo();
+        await context.driver.destroy();
+        context.driver = null;
+        if (info.mode === 'launched') {
+          lines.push(
+            `this session: stopped Chrome pid ${info.pid} (profile "${info.profile}")`
+          );
+        } else if (info.mode === 'external') {
+          lines.push(
+            `this session: disconnected from ${info.host}:${info.cdpPort} (not stopped; Pilot did not launch it)`
+          );
+        }
+      }
+      for (const result of await stopPilotBrowsers()) {
+        lines.push(`${result.profile}: ${result.outcome} - ${result.detail}`);
+      }
+      return {
+        content: [
+          {
+            type: 'text' as const,
+            text:
+              lines.length > 0
+                ? lines.join('\n')
+                : 'No Pilot-managed browsers were running.'
+          }
+        ]
+      };
+    }
+  );
+
   server.tool(
     'browser_screenshot',
     'Capture a screenshot of the current browser viewport or full page',
     screenshotShape,
     async ({ format, quality, fullPage }) => {
-      const { client } = requireClient(context);
-      const result = await executeScreenshot(client, {
+      const result = await requireDriver(context).screenshot({
         format,
         quality,
         fullPage
@@ -377,8 +415,7 @@ export function registerBrowserTools(
     'Navigate to a URL and wait for page load',
     navigateShape,
     async ({ url, waitUntil, timeoutMs }) => {
-      const { client, manager } = requireClient(context);
-      const result = await executeNavigate(client, manager, {
+      const result = await requireDriver(context).navigate({
         url,
         waitUntil,
         timeoutMs
@@ -416,8 +453,7 @@ export function registerBrowserTools(
         };
       }
 
-      const { client } = requireClient(context);
-      const result = await executeClick(client, params, context.elementRefMap);
+      const result = await requireDriver(context).click(params);
 
       return {
         content: [
@@ -435,8 +471,7 @@ export function registerBrowserTools(
     'Type text into the focused element, a ref (from browser_snapshot), or a CSS selector. Generates keystroke events (use browser_fill to replace a field value entirely).',
     typeShape,
     async (params) => {
-      const { client } = requireClient(context);
-      await executeType(client, params, context.elementRefMap);
+      await requireDriver(context).type(params);
 
       const target = params.ref ?? params.selector;
       return {
@@ -455,8 +490,7 @@ export function registerBrowserTools(
     'Execute JavaScript in the page context and return the result',
     evaluateShape,
     async ({ expression, awaitPromise, timeoutMs }) => {
-      const { client } = requireClient(context);
-      const result = await executeEvaluate(client, {
+      const result = await requireDriver(context).evaluate({
         expression,
         awaitPromise,
         timeoutMs
@@ -499,8 +533,7 @@ export function registerBrowserTools(
         };
       }
 
-      const { client, manager } = requireClient(context);
-      const result = await executeWait(client, manager, params);
+      const result = await requireDriver(context).wait(params);
 
       return {
         content: [
@@ -518,8 +551,7 @@ export function registerBrowserTools(
     'Get information about the current page (URL, title, ready state)',
     {},
     async () => {
-      const { client } = requireClient(context);
-      const result = await executePageInfo(client);
+      const result = await requireDriver(context).pageInfo();
 
       return {
         content: [
@@ -537,8 +569,7 @@ export function registerBrowserTools(
     'List all open browser tabs with their target IDs, URLs, titles, and active status',
     {},
     async () => {
-      const { manager } = requireClient(context);
-      const tabs = await executeListTabs(manager);
+      const tabs = await requireDriver(context).listTabs();
 
       return {
         content: [
@@ -556,8 +587,7 @@ export function registerBrowserTools(
     'Open a new browser tab, optionally navigating to a URL',
     newTabShape,
     async ({ url }) => {
-      const { manager } = requireClient(context);
-      const result = await executeNewTab(manager, url);
+      const result = await requireDriver(context).newTab(url);
 
       return {
         content: [
@@ -575,8 +605,7 @@ export function registerBrowserTools(
     'Close a browser tab by its target ID. Cannot close the currently active tab.',
     closeTabShape,
     async ({ targetId }) => {
-      const { manager } = requireClient(context);
-      await executeCloseTab(manager, targetId);
+      await requireDriver(context).closeTab(targetId);
 
       return {
         content: [
@@ -594,8 +623,7 @@ export function registerBrowserTools(
     'Switch the active browser tab to the one with the given target ID. All subsequent commands will target this tab.',
     switchTabShape,
     async ({ targetId }) => {
-      const { manager } = requireClient(context);
-      await executeSwitchTab(manager, targetId);
+      await requireDriver(context).switchTab(targetId);
 
       return {
         content: [
@@ -613,22 +641,12 @@ export function registerBrowserTools(
     'Resize the browser viewport to new dimensions. Updates both the window size and the rendering viewport. Disables responsive mode so the viewport stays locked at the specified size.',
     viewportResizeShape,
     async ({ width, height, deviceScaleFactor, mobile }) => {
-      const { client, manager } = requireClient(context);
-
-      const current = manager.getViewportConfig();
-      const effectiveMobile = mobile ?? current?.mobile ?? false;
-      const newConfig: ViewportConfig = {
+      const newConfig = await requireDriver(context).resizeViewport({
         width,
         height,
-        deviceScaleFactor: deviceScaleFactor ?? current?.deviceScaleFactor ?? 1,
-        mobile: effectiveMobile,
-        responsive: false,
-        ...(effectiveMobile &&
-          current?.userAgent && { userAgent: current.userAgent })
-      };
-
-      await applyViewport(client, newConfig);
-      manager.setViewportConfig(newConfig);
+        deviceScaleFactor,
+        mobile
+      });
 
       return {
         content: [
@@ -646,11 +664,11 @@ export function registerBrowserTools(
     'Manage browser tabs. Actions: "list" shows all tabs, "new" opens a tab (optional url), "select" switches to a tab (requires targetId), "close" closes a tab (requires targetId).',
     tabsShape,
     async ({ action, targetId, url }) => {
-      const { manager } = requireClient(context);
+      const driver = requireDriver(context);
 
       switch (action) {
         case 'list': {
-          const tabs = await executeListTabs(manager);
+          const tabs = await driver.listTabs();
           return {
             content: [
               { type: 'text' as const, text: JSON.stringify(tabs, null, 2) }
@@ -659,7 +677,7 @@ export function registerBrowserTools(
         }
 
         case 'new': {
-          const result = await executeNewTab(manager, url);
+          const result = await driver.newTab(url);
           return {
             content: [
               {
@@ -682,7 +700,7 @@ export function registerBrowserTools(
               isError: true
             };
           }
-          await executeSwitchTab(manager, targetId);
+          await driver.switchTab(targetId);
           return {
             content: [
               {
@@ -705,7 +723,7 @@ export function registerBrowserTools(
               isError: true
             };
           }
-          await executeCloseTab(manager, targetId);
+          await driver.closeTab(targetId);
           return {
             content: [
               {

package/src/mcp/tools/connect.ts

@@ -1,7 +1,6 @@
 import { z } from 'zod';
 import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
-import type { BrowserContext } from '../server.js';
-import type { BrowserManager } from '../../browser/browser-manager.js';
+import type { BrowserContext, DriverFactory } from '../server.js';
 import { executeConnect, formatConnectResult } from '../../commands/connect.js';
 
 const connectShape = {
@@ -21,14 +20,14 @@ const connectShape = {
 export function registerConnectTools(
   server: McpServer,
   context: BrowserContext,
-  makeBrowserManager: () => BrowserManager
+  makeDriver: DriverFactory
 ): void {
   server.tool(
     'browser_connect',
     'Connect to an existing Chrome instance with remote debugging enabled. Start Chrome with --remote-debugging-port=PORT, then use this tool to connect. Lists available tabs after connecting. Use browser_switch_tab to select a specific tab.',
     connectShape,
     async ({ host, port }) => {
-      if (context.manager?.isConnected()) {
+      if (context.driver?.isConnected()) {
         return {
           content: [
             {
@@ -39,22 +38,22 @@ export function registerConnectTools(
         };
       }
 
-      if (context.manager) {
-        await context.manager.destroy();
-        context.manager = null;
+      if (context.driver) {
+        await context.driver.destroy();
+        context.driver = null;
       }
 
-      const manager = makeBrowserManager();
+      const driver = makeDriver();
       try {
-        const result = await executeConnect(manager, { host, port });
-        context.manager = manager;
+        const result = await executeConnect(driver, { host, port });
+        context.driver = driver;
         return {
           content: [
             { type: 'text' as const, text: formatConnectResult(result) }
           ]
         };
       } catch (err) {
-        await manager.destroy();
+        await driver.destroy();
         throw err;
       }
     }

package/src/mcp/tools/driver-context.ts

@@ -0,0 +1,56 @@
+import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
+import {
+  TOOL_CAPABILITIES,
+  assertToolSupported
+} from '../../browser/driver/capabilities.js';
+import type { BrowserDriver } from '../../browser/driver/types.js';
+import type { ToolName } from './names.js';
+
+/** Return the connected driver or fail with the standard "not started" error. */
+export function requireDriver(context: {
+  driver: BrowserDriver | null;
+}): BrowserDriver {
+  const driver = context.driver;
+  if (!driver || !driver.isConnected()) {
+    throw new Error('Browser not started. Call browser_start first.');
+  }
+  return driver;
+}
+
+function isToolName(name: string): name is ToolName {
+  return Object.prototype.hasOwnProperty.call(TOOL_CAPABILITIES, name);
+}
+
+/**
+ * Wrap an McpServer so every registered tool first checks that the active
+ * driver (if any) has the capabilities the tool declares in TOOL_CAPABILITIES.
+ * Registering a tool without a capability declaration throws, so new tools
+ * cannot skip the check.
+ */
+export function withCapabilityGate(
+  server: McpServer,
+  getDriver: () => BrowserDriver | null
+): McpServer {
+  return new Proxy(server, {
+    get(target, prop, receiver) {
+      if (prop !== 'tool') return Reflect.get(target, prop, receiver);
+      return (name: string, ...rest: unknown[]) => {
+        if (!isToolName(name)) {
+          throw new Error(
+            `Tool ${name} has no entry in TOOL_CAPABILITIES; declare the driver capabilities it needs.`
+          );
+        }
+        const last = rest.length - 1;
+        const handler = rest[last];
+        if (typeof handler === 'function') {
+          rest[last] = async (...args: unknown[]) => {
+            const driver = getDriver();
+            if (driver) assertToolSupported(name, driver);
+            return handler(...args);
+          };
+        }
+        return (target.tool as (...a: unknown[]) => unknown)(name, ...rest);
+      };
+    }
+  });
+}

package/src/mcp/tools/files.ts

@@ -1,8 +1,8 @@
 import { z } from 'zod';
 import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
 import type { BrowserContext } from '../server.js';
-import { executeDialog, formatDialogResult } from '../../commands/dialog.js';
-import { executeUpload } from '../../commands/upload.js';
+import { formatDialogResult } from '../../commands/dialog.js';
+import { requireDriver } from './driver-context.js';
 import {
   executeDownloads,
   formatDownloadsResult
@@ -49,15 +49,6 @@ const downloadsShape = {
     .describe('Timeout for wait action in milliseconds (default: 30000)')
 };
 
-function requireContext(context: BrowserContext) {
-  const manager = context.manager;
-  if (!manager || !manager.isConnected()) {
-    throw new Error('Browser not started. Call browser_start first.');
-  }
-  const client = manager.getClient()!;
-  return { client };
-}
-
 export function registerFilesTools(
   server: McpServer,
   context: BrowserContext
@@ -67,8 +58,7 @@ export function registerFilesTools(
     'Handle browser dialogs (alert, confirm, prompt). Use "list" to see pending dialogs, "accept" to click OK/Yes, or "dismiss" to click Cancel/No. For prompt dialogs, provide promptText with accept.',
     dialogShape,
     async ({ action, promptText }) => {
-      const { client } = requireContext(context);
-      const result = await executeDialog(client, context.dialogQueue, {
+      const result = await requireDriver(context).dialog({
         action,
         promptText
       });
@@ -83,8 +73,7 @@ export function registerFilesTools(
     'Upload files to a file input element by ref. Files must exist, be regular files, and be under allowed roots (PILOT_UPLOAD_ROOTS env, default: cwd). Use browser_snapshot to find file input refs.',
     uploadShape,
     async ({ ref, files }) => {
-      const { client } = requireContext(context);
-      const result = await executeUpload(client, context.elementRefMap, {
+      const result = await requireDriver(context).upload({
         ref,
         files
       });
@@ -104,7 +93,7 @@ export function registerFilesTools(
     'Track file downloads. Actions: "list" shows tracked downloads with status/progress; "wait" blocks until a download (by GUID) completes; "clear" resets the tracker.',
     downloadsShape,
     async ({ action, guid, timeoutMs }) => {
-      requireContext(context);
+      requireDriver(context);
       const result = await executeDownloads(context.downloadTracker, {
         action,
         guid,

package/src/mcp/tools/interaction.ts

@@ -1,21 +1,7 @@
 import { z } from 'zod';
 import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
 import type { BrowserContext } from '../server.js';
-import { executeFill } from '../../commands/fill.js';
-import { executeHover } from '../../commands/hover.js';
-import { executePressKey } from '../../commands/press_key.js';
-import { executeSelect } from '../../commands/select.js';
-import { executeCheck } from '../../commands/check.js';
-import { executeScroll } from '../../commands/scroll.js';
-
-function requireContext(context: BrowserContext) {
-  const manager = context.manager;
-  if (!manager || !manager.isConnected()) {
-    throw new Error('Browser not started. Call browser_start first.');
-  }
-  const client = manager.getClient()!;
-  return { client, refMap: context.elementRefMap };
-}
+import { requireDriver } from './driver-context.js';
 
 const fillShape = {
   ref: z.string().describe('Element ref from browser_snapshot (e.g. "e3")'),
@@ -81,8 +67,7 @@ export function registerInteractionTools(
     'Fill a form field by element ref. Replaces the entire field value and fires input/change events compatible with React, Vue, and Svelte. Use browser_snapshot first to get refs.',
     fillShape,
     async ({ ref, value }) => {
-      const { client, refMap } = requireContext(context);
-      const result = await executeFill(client, refMap, { ref, value });
+      const result = await requireDriver(context).fill({ ref, value });
       return {
         content: [
           {
@@ -113,8 +98,7 @@ export function registerInteractionTools(
           isError: true
         };
       }
-      const { client, refMap } = requireContext(context);
-      const result = await executeHover(client, refMap, params);
+      const result = await requireDriver(context).hover(params);
       return {
         content: [
           {
@@ -131,8 +115,7 @@ export function registerInteractionTools(
     'Press a key or key combination. Supports named keys (Enter, Tab, Escape, etc.) and modifier combos (Control+a, Meta+c, Shift+Enter).',
     pressKeyShape,
     async ({ key }) => {
-      const { client } = requireContext(context);
-      const result = await executePressKey(client, { key });
+      const result = await requireDriver(context).pressKey({ key });
       const modStr =
         result.modifiers.length > 0
           ? ` with modifiers: ${result.modifiers.join('+')}`
@@ -165,8 +148,7 @@ export function registerInteractionTools(
           isError: true
         };
       }
-      const { client, refMap } = requireContext(context);
-      const result = await executeSelect(client, refMap, params);
+      const result = await requireDriver(context).select(params);
       return {
         content: [
           {
@@ -183,8 +165,7 @@ export function registerInteractionTools(
     'Toggle a checkbox or radio button by ref. Optionally set to a specific checked state.',
     checkShape,
     async ({ ref, checked }) => {
-      const { client, refMap } = requireContext(context);
-      const result = await executeCheck(client, refMap, { ref, checked });
+      const result = await requireDriver(context).check({ ref, checked });
       return {
         content: [
           {
@@ -217,8 +198,7 @@ export function registerInteractionTools(
           isError: true
         };
       }
-      const { client, refMap } = requireContext(context);
-      const result = await executeScroll(client, refMap, params);
+      const result = await requireDriver(context).scroll(params);
       return {
         content: [
           {

package/src/mcp/tools/names.ts

@@ -5,6 +5,8 @@
 export const EXPECTED_TOOLS = [
   'browser_start',
   'browser_stop',
+  'browser_status',
+  'browser_stop_all',
   'browser_screenshot',
   'browser_navigate',
   'browser_click',

package/src/mcp/tools/network.ts

@@ -1,8 +1,8 @@
 import { z } from 'zod';
 import type { McpServer } from '@modelcontextprotocol/sdk/server/mcp.js';
 import type { BrowserContext } from '../server.js';
+import { requireDriver } from './driver-context.js';
 import {
-  executeNetwork,
   formatNetworkList,
   formatNetworkDetail
 } from '../../commands/network.js';
@@ -48,15 +48,6 @@ const networkShape = {
     .describe('Maximum number of results (list only)')
 };
 
-function requireContext(context: BrowserContext) {
-  const manager = context.manager;
-  if (!manager || !manager.isConnected()) {
-    throw new Error('Browser not started. Call browser_start first.');
-  }
-  const client = manager.getClient()!;
-  return { client, networkBuffer: context.networkBuffer };
-}
-
 export function registerNetworkTools(
   server: McpServer,
   context: BrowserContext
@@ -66,8 +57,7 @@ export function registerNetworkTools(
     'Inspect captured network requests. Actions: "list" shows request summaries with filters (url, method, status, resourceType, failed); "get" returns full detail for a request ID including headers and response body; "clear" resets the buffer.',
     networkShape,
     async (params) => {
-      const { client, networkBuffer } = requireContext(context);
-      const result = await executeNetwork(client, networkBuffer, params);
+      const result = await requireDriver(context).network(params);
 
       let text: string;
       if (result.action === 'list') {