@ejazullah/browser-mcp 0.0.56

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (66) hide show
  1. package/LICENSE +202 -0
  2. package/README.md +860 -0
  3. package/cli.js +19 -0
  4. package/index.d.ts +23 -0
  5. package/index.js +1061 -0
  6. package/lib/auth.js +82 -0
  7. package/lib/browserContextFactory.js +205 -0
  8. package/lib/browserServerBackend.js +125 -0
  9. package/lib/config.js +266 -0
  10. package/lib/context.js +232 -0
  11. package/lib/databaseLogger.js +264 -0
  12. package/lib/extension/cdpRelay.js +346 -0
  13. package/lib/extension/extensionContextFactory.js +56 -0
  14. package/lib/extension/main.js +26 -0
  15. package/lib/fileUtils.js +32 -0
  16. package/lib/httpServer.js +39 -0
  17. package/lib/index.js +39 -0
  18. package/lib/javascript.js +49 -0
  19. package/lib/log.js +21 -0
  20. package/lib/loop/loop.js +69 -0
  21. package/lib/loop/loopClaude.js +152 -0
  22. package/lib/loop/loopOpenAI.js +143 -0
  23. package/lib/loop/main.js +60 -0
  24. package/lib/loopTools/context.js +66 -0
  25. package/lib/loopTools/main.js +49 -0
  26. package/lib/loopTools/perform.js +32 -0
  27. package/lib/loopTools/snapshot.js +29 -0
  28. package/lib/loopTools/tool.js +18 -0
  29. package/lib/manualPromise.js +111 -0
  30. package/lib/mcp/inProcessTransport.js +72 -0
  31. package/lib/mcp/server.js +93 -0
  32. package/lib/mcp/transport.js +217 -0
  33. package/lib/mongoDBLogger.js +252 -0
  34. package/lib/package.js +20 -0
  35. package/lib/program.js +113 -0
  36. package/lib/response.js +172 -0
  37. package/lib/sessionLog.js +156 -0
  38. package/lib/tab.js +266 -0
  39. package/lib/tools/cdp.js +169 -0
  40. package/lib/tools/common.js +55 -0
  41. package/lib/tools/console.js +33 -0
  42. package/lib/tools/dialogs.js +47 -0
  43. package/lib/tools/evaluate.js +53 -0
  44. package/lib/tools/extraction.js +217 -0
  45. package/lib/tools/files.js +44 -0
  46. package/lib/tools/forms.js +180 -0
  47. package/lib/tools/getext.js +99 -0
  48. package/lib/tools/install.js +53 -0
  49. package/lib/tools/interactions.js +191 -0
  50. package/lib/tools/keyboard.js +86 -0
  51. package/lib/tools/mouse.js +99 -0
  52. package/lib/tools/navigate.js +70 -0
  53. package/lib/tools/network.js +41 -0
  54. package/lib/tools/pdf.js +40 -0
  55. package/lib/tools/screenshot.js +75 -0
  56. package/lib/tools/selectors.js +233 -0
  57. package/lib/tools/snapshot.js +169 -0
  58. package/lib/tools/states.js +147 -0
  59. package/lib/tools/tabs.js +87 -0
  60. package/lib/tools/tool.js +33 -0
  61. package/lib/tools/utils.js +74 -0
  62. package/lib/tools/wait.js +56 -0
  63. package/lib/tools.js +64 -0
  64. package/lib/utils.js +26 -0
  65. package/openapi.json +683 -0
  66. package/package.json +92 -0
package/lib/context.js ADDED
@@ -0,0 +1,232 @@
1
+ /**
2
+ * Copyright (c) Microsoft Corporation.
3
+ *
4
+ * Licensed under the Apache License, Version 2.0 (the "License");
5
+ * you may not use this file except in compliance with the License.
6
+ * You may obtain a copy of the License at
7
+ *
8
+ * http://www.apache.org/licenses/LICENSE-2.0
9
+ *
10
+ * Unless required by applicable law or agreed to in writing, software
11
+ * distributed under the License is distributed on an "AS IS" BASIS,
12
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13
+ * See the License for the specific language governing permissions and
14
+ * limitations under the License.
15
+ */
16
+ import debug from 'debug';
17
+ import { logUnhandledError } from './log.js';
18
+ import { Tab } from './tab.js';
19
+ import { outputFile } from './config.js';
20
+ const testDebug = debug('pw:mcp:test');
21
+ export class Context {
22
+ tools;
23
+ config;
24
+ sessionLog;
25
+ options;
26
+ _browserContextPromise;
27
+ _browserContextFactory;
28
+ _tabs = [];
29
+ _currentTab;
30
+ _clientInfo;
31
+ static _allContexts = new Set();
32
+ _closeBrowserContextPromise;
33
+ _isRunningTool = false;
34
+ _abortController = new AbortController();
35
+ constructor(options) {
36
+ this.tools = options.tools;
37
+ this.config = options.config;
38
+ this.sessionLog = options.sessionLog;
39
+ this.options = options;
40
+ this._browserContextFactory = options.browserContextFactory;
41
+ this._clientInfo = options.clientInfo;
42
+ testDebug('create context');
43
+ Context._allContexts.add(this);
44
+ }
45
+ static async disposeAll() {
46
+ await Promise.all([...Context._allContexts].map(context => context.dispose()));
47
+ }
48
+ tabs() {
49
+ return this._tabs;
50
+ }
51
+ currentTab() {
52
+ return this._currentTab;
53
+ }
54
+ currentTabOrDie() {
55
+ if (!this._currentTab)
56
+ throw new Error('No open pages available. Use the "browser_navigate" tool to navigate to a page first.');
57
+ return this._currentTab;
58
+ }
59
+ async newTab() {
60
+ const { browserContext } = await this._ensureBrowserContext();
61
+ const page = await browserContext.newPage();
62
+ this._currentTab = this._tabs.find(t => t.page === page);
63
+ return this._currentTab;
64
+ }
65
+ async selectTab(index) {
66
+ const tab = this._tabs[index];
67
+ if (!tab)
68
+ throw new Error(`Tab ${index} not found`);
69
+ await tab.page.bringToFront();
70
+ this._currentTab = tab;
71
+ return tab;
72
+ }
73
+ async updateBrowserContextFactory(newFactory) {
74
+ // Close existing browser context first
75
+ await this.closeBrowserContext();
76
+ // Update to the new factory
77
+ this._browserContextFactory = newFactory;
78
+ }
79
+ async ensureTab() {
80
+ const { browserContext } = await this._ensureBrowserContext();
81
+ if (!this._currentTab)
82
+ await browserContext.newPage();
83
+ return this._currentTab;
84
+ }
85
+ async closeTab(index) {
86
+ const tab = index === undefined ? this._currentTab : this._tabs[index];
87
+ if (!tab)
88
+ throw new Error(`Tab ${index} not found`);
89
+ const url = tab.page.url();
90
+ await tab.page.close();
91
+ return url;
92
+ }
93
+ async outputFile(name) {
94
+ return outputFile(this.config, this._clientInfo.rootPath, name);
95
+ }
96
+ _onPageCreated(page) {
97
+ const tab = new Tab(this, page, tab => this._onPageClosed(tab));
98
+ this._tabs.push(tab);
99
+ if (!this._currentTab)
100
+ this._currentTab = tab;
101
+ }
102
+ _onPageClosed(tab) {
103
+ const index = this._tabs.indexOf(tab);
104
+ if (index === -1)
105
+ return;
106
+ this._tabs.splice(index, 1);
107
+ if (this._currentTab === tab)
108
+ this._currentTab = this._tabs[Math.min(index, this._tabs.length - 1)];
109
+ if (!this._tabs.length)
110
+ void this.closeBrowserContext();
111
+ }
112
+ async closeBrowserContext() {
113
+ if (!this._closeBrowserContextPromise)
114
+ this._closeBrowserContextPromise = this._closeBrowserContextImpl().catch(logUnhandledError);
115
+ await this._closeBrowserContextPromise;
116
+ this._closeBrowserContextPromise = undefined;
117
+ }
118
+ isRunningTool() {
119
+ return this._isRunningTool;
120
+ }
121
+ setRunningTool(isRunningTool) {
122
+ this._isRunningTool = isRunningTool;
123
+ }
124
+ async _closeBrowserContextImpl() {
125
+ if (!this._browserContextPromise)
126
+ return;
127
+ testDebug('close context');
128
+ const promise = this._browserContextPromise;
129
+ this._browserContextPromise = undefined;
130
+ await promise.then(async ({ browserContext, close }) => {
131
+ if (this.config.saveTrace)
132
+ await browserContext.tracing.stop();
133
+ await close();
134
+ });
135
+ }
136
+ async dispose() {
137
+ this._abortController.abort('MCP context disposed');
138
+ await this.closeBrowserContext();
139
+ Context._allContexts.delete(this);
140
+ }
141
+ async _setupRequestInterception(context) {
142
+ if (this.config.network?.allowedOrigins?.length) {
143
+ await context.route('**', route => route.abort('blockedbyclient'));
144
+ for (const origin of this.config.network.allowedOrigins)
145
+ await context.route(`*://${origin}/**`, route => route.continue());
146
+ }
147
+ if (this.config.network?.blockedOrigins?.length) {
148
+ for (const origin of this.config.network.blockedOrigins)
149
+ await context.route(`*://${origin}/**`, route => route.abort('blockedbyclient'));
150
+ }
151
+ }
152
+ _ensureBrowserContext() {
153
+ if (!this._browserContextPromise) {
154
+ this._browserContextPromise = this._setupBrowserContext();
155
+ this._browserContextPromise.catch(() => {
156
+ this._browserContextPromise = undefined;
157
+ });
158
+ }
159
+ return this._browserContextPromise;
160
+ }
161
+ async _setupBrowserContext() {
162
+ if (this._closeBrowserContextPromise)
163
+ throw new Error('Another browser context is being closed.');
164
+ // TODO: move to the browser context factory to make it based on isolation mode.
165
+ const result = await this._browserContextFactory.createContext(this._clientInfo, this._abortController.signal);
166
+ const { browserContext } = result;
167
+ await this._setupRequestInterception(browserContext);
168
+ if (this.sessionLog)
169
+ await InputRecorder.create(this, browserContext);
170
+ for (const page of browserContext.pages())
171
+ this._onPageCreated(page);
172
+ browserContext.on('page', page => this._onPageCreated(page));
173
+ if (this.config.saveTrace) {
174
+ await browserContext.tracing.start({
175
+ name: 'trace',
176
+ screenshots: false,
177
+ snapshots: true,
178
+ sources: false,
179
+ });
180
+ }
181
+ return result;
182
+ }
183
+ }
184
+ export class InputRecorder {
185
+ _context;
186
+ _browserContext;
187
+ constructor(context, browserContext) {
188
+ this._context = context;
189
+ this._browserContext = browserContext;
190
+ }
191
+ static async create(context, browserContext) {
192
+ const recorder = new InputRecorder(context, browserContext);
193
+ await recorder._initialize();
194
+ return recorder;
195
+ }
196
+ async _initialize() {
197
+ const sessionLog = this._context.sessionLog;
198
+ await this._browserContext._enableRecorder({
199
+ mode: 'recording',
200
+ recorderMode: 'api',
201
+ }, {
202
+ actionAdded: (page, data, code) => {
203
+ if (this._context.isRunningTool())
204
+ return;
205
+ const tab = Tab.forPage(page);
206
+ if (tab)
207
+ sessionLog.logUserAction(data.action, tab, code, false);
208
+ },
209
+ actionUpdated: (page, data, code) => {
210
+ if (this._context.isRunningTool())
211
+ return;
212
+ const tab = Tab.forPage(page);
213
+ if (tab)
214
+ sessionLog.logUserAction(data.action, tab, code, true);
215
+ },
216
+ signalAdded: (page, data) => {
217
+ if (this._context.isRunningTool())
218
+ return;
219
+ if (data.signal.name !== 'navigation')
220
+ return;
221
+ const tab = Tab.forPage(page);
222
+ const navigateAction = {
223
+ name: 'navigate',
224
+ url: data.signal.url,
225
+ signals: [],
226
+ };
227
+ if (tab)
228
+ sessionLog.logUserAction(navigateAction, tab, `await page.goto('${data.signal.url}');`, false);
229
+ },
230
+ });
231
+ }
232
+ }
@@ -0,0 +1,264 @@
1
+ /**
2
+ * Copyright (c) Microsoft Corporation.
3
+ *
4
+ * Licensed under the Apache License, Version 2.0 (the "License");
5
+ * you may not use this file except in compliance with the License.
6
+ * You may obtain a copy of the License at
7
+ *
8
+ * http://www.apache.org/licenses/LICENSE-2.0
9
+ *
10
+ * Unless required by applicable law or agreed to in writing, software
11
+ * distributed under the License is distributed on an "AS IS" BASIS,
12
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13
+ * See the License for the specific language governing permissions and
14
+ * limitations under the License.
15
+ */
16
+ import fs from 'fs';
17
+ import path from 'path';
18
+ import { logUnhandledError } from './log.js';
19
+ import { MongoDBLogger } from './mongoDBLogger.js';
20
+ export class DatabaseLogger {
21
+ _dbFile;
22
+ _pendingWrites = [];
23
+ _writeQueue = Promise.resolve();
24
+ _flushTimeout;
25
+ _mongoLogger = null;
26
+ _useMongo = false;
27
+ _sessionId;
28
+ constructor(dbFilePath, mongoConfig) {
29
+ this._dbFile = dbFilePath;
30
+ this._sessionId = mongoConfig?.sessionId || `session_${Date.now()}`;
31
+ this._initializeDatabase();
32
+ // Try to initialize MongoDB if config is provided or env var is set
33
+ if (mongoConfig || process.env.MONGODB_URL) {
34
+ this._initializeMongoDB(mongoConfig).catch(logUnhandledError);
35
+ }
36
+ }
37
+ _initializeDatabase() {
38
+ // Create database file with header if it doesn't exist
39
+ if (!fs.existsSync(this._dbFile)) {
40
+ const header = [];
41
+ fs.writeFileSync(this._dbFile, JSON.stringify(header, null, 2));
42
+ }
43
+ }
44
+ async _initializeMongoDB(config) {
45
+ try {
46
+ this._mongoLogger = new MongoDBLogger({ ...config, sessionId: this._sessionId });
47
+ const connected = await this._mongoLogger.connect();
48
+ if (connected) {
49
+ this._useMongo = true;
50
+ // eslint-disable-next-line no-console
51
+ console.error(`✓ Using MongoDB for interaction logging (Session: ${this._sessionId})`);
52
+ }
53
+ else {
54
+ // eslint-disable-next-line no-console
55
+ console.error('✓ Using JSON file for interaction logging (MongoDB unavailable)');
56
+ }
57
+ }
58
+ catch (error) {
59
+ // eslint-disable-next-line no-console
60
+ console.error('✓ Using JSON file for interaction logging (MongoDB unavailable)');
61
+ logUnhandledError(error);
62
+ }
63
+ }
64
+ /**
65
+ * Log an element interaction to the database
66
+ */
67
+ logInteraction(interaction) {
68
+ const interactionWithSession = { ...interaction, sessionId: this._sessionId };
69
+ // Log to MongoDB if available
70
+ if (this._useMongo && this._mongoLogger) {
71
+ this._mongoLogger.logInteraction(interactionWithSession).catch(logUnhandledError);
72
+ }
73
+ // Always log to JSON file as backup
74
+ this._pendingWrites.push(interactionWithSession);
75
+ // Debounce writes
76
+ if (this._flushTimeout)
77
+ clearTimeout(this._flushTimeout);
78
+ this._flushTimeout = setTimeout(() => this._flush(), 500);
79
+ }
80
+ async _flush() {
81
+ if (this._pendingWrites.length === 0)
82
+ return;
83
+ const interactions = [...this._pendingWrites];
84
+ this._pendingWrites = [];
85
+ this._writeQueue = this._writeQueue
86
+ .then(async () => {
87
+ try {
88
+ // Read existing data
89
+ let existingData = [];
90
+ if (fs.existsSync(this._dbFile)) {
91
+ const content = await fs.promises.readFile(this._dbFile, 'utf-8');
92
+ if (content.trim()) {
93
+ existingData = JSON.parse(content);
94
+ }
95
+ }
96
+ // Append new interactions
97
+ existingData.push(...interactions);
98
+ // Write back to file
99
+ await fs.promises.writeFile(this._dbFile, JSON.stringify(existingData, null, 2), 'utf-8');
100
+ }
101
+ catch (error) {
102
+ logUnhandledError(error);
103
+ }
104
+ })
105
+ .catch(logUnhandledError);
106
+ }
107
+ /**
108
+ * Flush all pending writes immediately
109
+ */
110
+ async flushSync() {
111
+ if (this._flushTimeout) {
112
+ clearTimeout(this._flushTimeout);
113
+ this._flushTimeout = undefined;
114
+ }
115
+ await this._flush();
116
+ await this._writeQueue;
117
+ // Flush MongoDB if available
118
+ if (this._mongoLogger) {
119
+ await this._mongoLogger.flushSync();
120
+ }
121
+ }
122
+ /**
123
+ * Get MongoDB logger instance
124
+ */
125
+ getMongoLogger() {
126
+ return this._mongoLogger;
127
+ }
128
+ /**
129
+ * Check if using MongoDB
130
+ */
131
+ isUsingMongo() {
132
+ return this._useMongo;
133
+ }
134
+ /**
135
+ * Get session ID
136
+ */
137
+ getSessionId() {
138
+ return this._sessionId;
139
+ }
140
+ /**
141
+ * Disconnect from MongoDB
142
+ */
143
+ async disconnect() {
144
+ if (this._mongoLogger) {
145
+ await this._mongoLogger.disconnect();
146
+ }
147
+ }
148
+ /**
149
+ * Query interactions by tool name
150
+ */
151
+ async queryByToolName(toolName, sessionId) {
152
+ await this.flushSync();
153
+ // Try MongoDB first
154
+ if (this._useMongo && this._mongoLogger) {
155
+ return await this._mongoLogger.queryByToolName(toolName, sessionId);
156
+ }
157
+ // Fall back to JSON file
158
+ if (!fs.existsSync(this._dbFile))
159
+ return [];
160
+ const content = await fs.promises.readFile(this._dbFile, 'utf-8');
161
+ if (!content.trim())
162
+ return [];
163
+ const data = JSON.parse(content);
164
+ return data.filter(item => item.toolName === toolName &&
165
+ (!sessionId || item.sessionId === sessionId || item.sessionId === this._sessionId));
166
+ }
167
+ /**
168
+ * Query interactions by element ref
169
+ */
170
+ async queryByElementRef(ref, sessionId) {
171
+ await this.flushSync();
172
+ // Try MongoDB first
173
+ if (this._useMongo && this._mongoLogger) {
174
+ return await this._mongoLogger.queryByElementRef(ref, sessionId);
175
+ }
176
+ // Fall back to JSON file
177
+ if (!fs.existsSync(this._dbFile))
178
+ return [];
179
+ const content = await fs.promises.readFile(this._dbFile, 'utf-8');
180
+ if (!content.trim())
181
+ return [];
182
+ const data = JSON.parse(content);
183
+ return data.filter(item => item.elementRef === ref &&
184
+ (!sessionId || item.sessionId === sessionId || item.sessionId === this._sessionId));
185
+ }
186
+ /**
187
+ * Query interactions by URL
188
+ */
189
+ async queryByUrl(url, sessionId) {
190
+ await this.flushSync();
191
+ // Try MongoDB first
192
+ if (this._useMongo && this._mongoLogger) {
193
+ return await this._mongoLogger.queryByUrl(url, sessionId);
194
+ }
195
+ // Fall back to JSON file
196
+ if (!fs.existsSync(this._dbFile))
197
+ return [];
198
+ const content = await fs.promises.readFile(this._dbFile, 'utf-8');
199
+ if (!content.trim())
200
+ return [];
201
+ const data = JSON.parse(content);
202
+ return data.filter(item => item.url.includes(url) &&
203
+ (!sessionId || item.sessionId === sessionId || item.sessionId === this._sessionId));
204
+ }
205
+ /**
206
+ * Get all interactions
207
+ */
208
+ async getAllInteractions(sessionId) {
209
+ await this.flushSync();
210
+ // Try MongoDB first
211
+ if (this._useMongo && this._mongoLogger) {
212
+ return await this._mongoLogger.getAllInteractions(sessionId);
213
+ }
214
+ // Fall back to JSON file
215
+ if (!fs.existsSync(this._dbFile))
216
+ return [];
217
+ const content = await fs.promises.readFile(this._dbFile, 'utf-8');
218
+ if (!content.trim())
219
+ return [];
220
+ const data = JSON.parse(content);
221
+ if (!sessionId)
222
+ return data.filter(item => item.sessionId === this._sessionId);
223
+ return data.filter(item => item.sessionId === sessionId);
224
+ }
225
+ /**
226
+ * Get interaction statistics
227
+ */
228
+ async getStatistics(sessionId) {
229
+ // Try MongoDB first
230
+ if (this._useMongo && this._mongoLogger) {
231
+ return await this._mongoLogger.getStatistics(sessionId);
232
+ }
233
+ // Fall back to calculating from JSON file
234
+ const interactions = await this.getAllInteractions(sessionId);
235
+ const toolCounts = {};
236
+ const elementCounts = {};
237
+ const urlCounts = {};
238
+ interactions.forEach(interaction => {
239
+ toolCounts[interaction.toolName] = (toolCounts[interaction.toolName] || 0) + 1;
240
+ elementCounts[interaction.elementRef] = (elementCounts[interaction.elementRef] || 0) + 1;
241
+ urlCounts[interaction.url] = (urlCounts[interaction.url] || 0) + 1;
242
+ });
243
+ return {
244
+ totalInteractions: interactions.length,
245
+ toolCounts,
246
+ elementCounts,
247
+ urlCounts,
248
+ };
249
+ }
250
+ /**
251
+ * Clear all data from the database
252
+ */
253
+ async clearDatabase() {
254
+ await this.flushSync();
255
+ await fs.promises.writeFile(this._dbFile, JSON.stringify([], null, 2));
256
+ }
257
+ }
258
+ /**
259
+ * Create a database logger instance
260
+ */
261
+ export function createDatabaseLogger(outputFolder, mongoConfig) {
262
+ const dbPath = path.join(outputFolder, 'interactions.json');
263
+ return new DatabaseLogger(dbPath, mongoConfig);
264
+ }