chrome-bridge-mcp 1.10.1 → 1.11.1

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.
package/README.md CHANGED
@@ -1,6 +1,6 @@
1
1
  # Chrome Bridge
2
2
 
3
- ![License: MIT](https://img.shields.io/badge/license-MIT-green) ![Node 18+](https://img.shields.io/badge/node-%E2%89%A518-brightgreen) ![Chrome 135+](https://img.shields.io/badge/chrome-%E2%89%A5135-blue) ![Tests](https://img.shields.io/badge/tests-132%20unit%20%2B%20e2e-brightgreen) [![Chrome Web Store](https://img.shields.io/badge/web%20store-published-blue)](https://chromewebstore.google.com/detail/chrome-bridge-for-claude/bioknpaeahidbelaljjohjofiloeodmb)
3
+ ![License: MIT](https://img.shields.io/badge/license-MIT-green) ![Node 18+](https://img.shields.io/badge/node-%E2%89%A518-brightgreen) ![Chrome 135+](https://img.shields.io/badge/chrome-%E2%89%A5135-blue) ![Tests](https://img.shields.io/badge/tests-147%20unit%20%2B%20e2e-brightgreen) [![Chrome Web Store](https://img.shields.io/badge/web%20store-published-blue)](https://chromewebstore.google.com/detail/chrome-bridge-for-claude/bioknpaeahidbelaljjohjofiloeodmb)
4
4
 
5
5
  **Chrome Bridge is an MCP server that connects Claude Code to your real, logged-in Chrome browser — measured 2.75× fewer turns and 2.28× lower cost than the official "Claude in Chrome" extension on a form-filling task, with ~3× the toolset and no paid plan.**
6
6
 
@@ -63,10 +63,10 @@ Claude Code calls `navigate` (returning clickable element refs), `accessibility_
63
63
 
64
64
  ## Token-Efficient Design
65
65
 
66
- - **Opt-in Surface**: 31 core tools cost ≈7.6k tokens of `tools/list`, all 60 cost ≈14.6k; specialized groups (audits, visual, network, storage, dom, files) are opt-in via `--caps`. That core figure was ≈3.8k in 1.8.0 and roughly doubled in 1.10.0: MCP annotations on every tool, rewritten descriptions, and a documented `.describe()` on all 254 parameters (coverage went from 35% to 100%). The trade is deliberate and it is not free — on the form benchmark it adds ≈8% to the cache-read tokens per turn — but the measured advantage was never prefix size, it was round trips: 2.75× fewer turns. Playwright MCP's 23 core tools measured ≈4.6k in July 2026, so on prefix size alone it is the leaner one. Measure ours with `npm run measure`.
66
+ - **Opt-in Surface**: 31 core tools cost ≈7.9k tokens of `tools/list`, all 60 cost ≈14.9k; specialized groups (audits, visual, network, storage, dom, files) are opt-in via `--caps`. That core figure was ≈3.8k in 1.8.0 and roughly doubled in 1.10.0: MCP annotations on every tool, rewritten descriptions, and a documented `.describe()` on all 254 parameters (coverage went from 35% to 100%). The trade is deliberate and it is not free — on the form benchmark it adds ≈8% to the cache-read tokens per turn — but the measured advantage was never prefix size, it was round trips: 2.75× fewer turns. Playwright MCP's 23 core tools measured ≈4.6k in July 2026, so on prefix size alone it is the leaner one. Measure ours with `npm run measure`.
67
67
  - **Act-from-Result**: Tools attach a capped preview of interactive elements with short refs. Actions only report a `page_changed` delta when the URL or title actually changes.
68
68
  - **Optimized Media**: Screenshots are downscaled to ≤1568px. Full-page captures are sliced into readable segments.
69
- - **Zero-Token Escape Hatches**: The [CLI](#cli) skips MCP schemas entirely, and recorded flows [replay](#launch-mode-headless--ci) without any model in the loop.
69
+ - **Payload Escape Hatches**: `save_to` on `read_page`, `extract`, `screenshot` and `http_request` writes the result to a file and returns the path — the bytes never enter the context unless the agent decides to read them. `read_page(mode="markdown")` keeps headings, links and tables at a fraction of the HTML cost. The [CLI](#cli) skips MCP schemas entirely, and recorded flows [replay](#launch-mode-headless--ci) without any model in the loop.
70
70
 
71
71
  ## Highlights
72
72
 
@@ -0,0 +1,151 @@
1
+ /**
2
+ * Rende una pagina in Markdown per il consumo del modello: struttura preservata
3
+ * a una frazione del costo dell'HTML.
4
+ *
5
+ * La funzione viene serializzata e iniettata nella pagina con
6
+ * chrome.scripting.executeScript, quindi deve essere autocontenuta: niente
7
+ * riferimenti a variabili esterne, niente import. Per lo stesso motivo è testata
8
+ * a parte contro un DOM finto (`test/unit/page-markdown.test.js`) invece che dal
9
+ * vivo.
10
+ *
11
+ * Cammina l'albero per blocchi invece di usare un querySelectorAll piatto: un
12
+ * `<li><a>x</a></li>` con il selettore piatto verrebbe emesso due volte, una
13
+ * come voce di lista e una come link.
14
+ */
15
+ export function buildMarkdown(doc) {
16
+ const d = doc || document;
17
+ const out = [];
18
+
19
+ const SKIP = new Set(['SCRIPT', 'STYLE', 'NOSCRIPT', 'TEMPLATE', 'SVG', 'CANVAS', 'IFRAME']);
20
+ const HEADINGS = new Set(['H1', 'H2', 'H3', 'H4', 'H5', 'H6']);
21
+ const BLOCKS = new Set(['P', 'LI', 'BLOCKQUOTE', 'PRE', 'DT', 'DD', 'FIGCAPTION']);
22
+ const MAX_TABLE_ROWS = 50;
23
+
24
+ const attr = (el, name) => (el.getAttribute ? el.getAttribute(name) : null);
25
+ const clean = (s) => String(s || '').replace(/\s+/g, ' ').trim();
26
+
27
+ // Dimensioni: il rect è la verità quando c'è un layout, ma su un'immagine non
28
+ // ancora disposta (o in test) restano solo gli attributi.
29
+ const sizeOf = (el) => {
30
+ let w = 0;
31
+ let h = 0;
32
+ if (typeof el.getBoundingClientRect === 'function') {
33
+ const r = el.getBoundingClientRect();
34
+ w = r.width; h = r.height;
35
+ }
36
+ if (!w) w = Number(attr(el, 'width')) || 0;
37
+ if (!h) h = Number(attr(el, 'height')) || 0;
38
+ return { w, h };
39
+ };
40
+
41
+ // Testo inline di un blocco: i link diventano [testo](href) senza che il
42
+ // testo venga emesso una seconda volta dal walker.
43
+ const inline = (node) => {
44
+ if (node.nodeType === 3) return node.textContent || '';
45
+ if (node.nodeType !== 1) return '';
46
+ const tag = node.tagName;
47
+ if (SKIP.has(tag)) return '';
48
+ if (tag === 'BR') return ' ';
49
+ if (tag === 'IMG') {
50
+ const { w, h } = sizeOf(node);
51
+ if (w >= 50 && h >= 50) {
52
+ const alt = clean(attr(node, 'alt')) || 'no alt text';
53
+ return ` ![${alt} ${Math.round(w)}x${Math.round(h)}](${attr(node, 'src') || ''}) `;
54
+ }
55
+ return '';
56
+ }
57
+ const kids = (node.childNodes || []).map(inline).join('');
58
+ if (tag === 'A') {
59
+ const href = attr(node, 'href');
60
+ const label = clean(kids);
61
+ if (!label) return '';
62
+ return href ? `[${label}](${href})` : label;
63
+ }
64
+ if (tag === 'CODE' || tag === 'KBD') {
65
+ const label = clean(kids);
66
+ return label ? '`' + label + '`' : '';
67
+ }
68
+ if (tag === 'STRONG' || tag === 'B') { const l = clean(kids); return l ? `**${l}**` : ''; }
69
+ if (tag === 'EM' || tag === 'I') { const l = clean(kids); return l ? `*${l}*` : ''; }
70
+ return kids;
71
+ };
72
+
73
+ const cellsOf = (row) => (row.childNodes || [])
74
+ .filter((n) => n.nodeType === 1 && (n.tagName === 'TD' || n.tagName === 'TH'))
75
+ .map((c) => clean(c.textContent).replace(/\|/g, '\\|'));
76
+
77
+ const renderTable = (table) => {
78
+ const rows = (table.querySelectorAll ? table.querySelectorAll('tr') : []) || [];
79
+ if (!rows.length) return;
80
+ const all = rows.map(cellsOf).filter((r) => r.length);
81
+ if (!all.length) return;
82
+
83
+ // Gli header veri, non un segnaposto: la prima riga se contiene th,
84
+ // altrimenti colonne numerate.
85
+ const firstHasTh = (rows[0].childNodes || []).some((n) => n.nodeType === 1 && n.tagName === 'TH');
86
+ const header = firstHasTh ? all[0] : all[0].map((_, i) => `col${i + 1}`);
87
+ const body = firstHasTh ? all.slice(1) : all;
88
+ const width = header.length;
89
+
90
+ out.push('');
91
+ out.push(`| ${header.join(' | ')} |`);
92
+ out.push(`|${' --- |'.repeat(width)}`);
93
+ for (const r of body.slice(0, MAX_TABLE_ROWS)) {
94
+ const padded = r.slice(0, width);
95
+ while (padded.length < width) padded.push('');
96
+ out.push(`| ${padded.join(' | ')} |`);
97
+ }
98
+ if (body.length > MAX_TABLE_ROWS) {
99
+ out.push(`_${body.length} rows total, ${body.length - MAX_TABLE_ROWS} omitted — use extract_table with \`where\` to filter server-side instead of reading them here._`);
100
+ }
101
+ out.push('');
102
+ };
103
+
104
+ const walk = (node) => {
105
+ if (!node || node.nodeType !== 1) return;
106
+ const tag = node.tagName;
107
+ if (SKIP.has(tag)) return;
108
+
109
+ if (tag === 'TABLE') { renderTable(node); return; }
110
+
111
+ if (HEADINGS.has(tag)) {
112
+ const text = clean(inline(node));
113
+ if (text) out.push(`${'#'.repeat(Number(tag[1]))} ${text}`, '');
114
+ return;
115
+ }
116
+
117
+ if (BLOCKS.has(tag)) {
118
+ const text = clean(inline(node));
119
+ if (!text) return;
120
+ if (tag === 'LI') out.push(`- ${text}`);
121
+ else if (tag === 'BLOCKQUOTE') out.push(`> ${text}`, '');
122
+ else if (tag === 'PRE') out.push('```', text, '```', '');
123
+ else if (tag === 'FIGCAPTION') out.push(`*${text}*`, '');
124
+ else out.push(text, '');
125
+ return;
126
+ }
127
+
128
+ if (tag === 'IMG') {
129
+ const rendered = clean(inline(node));
130
+ if (rendered) out.push(rendered, '');
131
+ return;
132
+ }
133
+
134
+ for (const child of node.childNodes || []) walk(child);
135
+ };
136
+
137
+ const title = clean(d.title);
138
+ if (title) out.push(`# ${title}`, '');
139
+
140
+ // Sommario delle immagini grandi: dice al modello se vale la pena chiedere
141
+ // uno screenshot, al costo di una riga.
142
+ const imgs = (d.body?.querySelectorAll ? d.body.querySelectorAll('img') : []) || [];
143
+ const significant = imgs.filter((img) => { const { w, h } = sizeOf(img); return w >= 100 && h >= 100; });
144
+ if (significant.length) {
145
+ out.push(`_${significant.length} image(s) at least 100x100 on this page — take a screenshot if the visual matters._`, '');
146
+ }
147
+
148
+ walk(d.body);
149
+
150
+ return out.join('\n').replace(/\n{3,}/g, '\n\n').trim();
151
+ }
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "manifest_version": 3,
3
3
  "name": "Chrome Bridge for Claude Code",
4
- "version": "1.10.1",
4
+ "version": "1.11.1",
5
5
  "description": "Bridge your browser to Claude Code: 60 token-efficient web-dev automation tools over a local WebSocket. ChromeOS included.",
6
6
  "homepage_url": "https://github.com/frsorrentino/chrome-bridge",
7
7
  "minimum_chrome_version": "135",
@@ -6,6 +6,7 @@
6
6
  */
7
7
 
8
8
  import './telemetry.js';
9
+ import { buildMarkdown } from './lib/page-markdown.js';
9
10
  const { pushError } = globalThis.__cbTelemetry;
10
11
 
11
12
  const DEFAULT_PORT = 8765;
@@ -423,6 +424,8 @@ async function executeCommand(msg) {
423
424
  return await cmdManageDownloads(params);
424
425
  case 'save_page':
425
426
  return await cmdSavePage(params);
427
+ case 'wait_for_text':
428
+ return await cmdWaitForText(params);
426
429
  case 'http_request':
427
430
  return await cmdHttpRequest(params);
428
431
  case 'set_zoom':
@@ -722,13 +725,13 @@ async function cmdExecuteJs({ code, tab_id, frame_id }) {
722
725
  }
723
726
  }
724
727
 
725
- async function cmdClick({ selector, tab_id, frame_id, force = false }) {
728
+ async function cmdClick({ selector, tab_id, frame_id, force = false, button = 'left', count = 1 }) {
726
729
  if (!selector) throw new Error('Missing required parameter: selector');
727
730
  const tabId = await resolveTabId(tab_id);
728
731
 
729
732
  const results = await chrome.scripting.executeScript({
730
733
  target: scriptTarget(tabId, frame_id),
731
- func: (sel, forceClick) => {
734
+ func: (sel, forceClick, btn, times) => {
732
735
  function deepQuery(sel) {
733
736
  if (!sel.includes('>>>')) return document.querySelector(sel);
734
737
  const parts = sel.split('>>>').map((s) => s.trim());
@@ -755,15 +758,29 @@ async function cmdClick({ selector, tab_id, frame_id, force = false }) {
755
758
  return { clicked: false, occluded: true, occluder: { selector: occluderSel, tag: top.tagName.toLowerCase(), text: (top.textContent || '').trim().substring(0, 80) } };
756
759
  }
757
760
  }
758
- const opts = { bubbles: true, cancelable: true, clientX: cx, clientY: cy };
759
- el.dispatchEvent(new PointerEvent('pointerdown', opts));
760
- el.dispatchEvent(new MouseEvent('mousedown', opts));
761
- el.dispatchEvent(new PointerEvent('pointerup', opts));
762
- el.dispatchEvent(new MouseEvent('mouseup', opts));
763
- el.click();
764
- return { clicked: true, tagName: el.tagName, text: el.textContent?.substring(0, 100) };
761
+ // button 2 = destro: il browser non apre il suo menu nativo da un evento
762
+ // sintetico, ma le pagine che ne implementano uno proprio ascoltano
763
+ // 'contextmenu', ed è quello il caso d'uso reale.
764
+ const right = btn === 'right';
765
+ const clickCount = Math.max(1, Number(times) || 1);
766
+ const base = { bubbles: true, cancelable: true, clientX: cx, clientY: cy, button: right ? 2 : 0, buttons: right ? 2 : 1 };
767
+
768
+ for (let i = 0; i < clickCount; i++) {
769
+ const opts = { ...base, detail: i + 1 };
770
+ el.dispatchEvent(new PointerEvent('pointerdown', opts));
771
+ el.dispatchEvent(new MouseEvent('mousedown', opts));
772
+ el.dispatchEvent(new PointerEvent('pointerup', opts));
773
+ el.dispatchEvent(new MouseEvent('mouseup', opts));
774
+ if (right) el.dispatchEvent(new MouseEvent('contextmenu', opts));
775
+ else el.click();
776
+ }
777
+ // dblclick non è implicito nei due click: va emesso a parte, ed è quello
778
+ // che seleziona una parola o apre un editor inline.
779
+ if (!right && clickCount >= 2) el.dispatchEvent(new MouseEvent('dblclick', { ...base, detail: 2 }));
780
+
781
+ return { clicked: true, button: right ? 'right' : 'left', count: clickCount, tagName: el.tagName, text: el.textContent?.substring(0, 100) };
765
782
  },
766
- args: [selector, force],
783
+ args: [selector, force, button, count],
767
784
  world: 'MAIN',
768
785
  });
769
786
 
@@ -853,6 +870,18 @@ async function cmdReadPage({ mode = 'text', tab_id, frame_id }) {
853
870
  return results?.[0]?.result ?? '';
854
871
  }
855
872
 
873
+ if (mode === 'markdown') {
874
+ // buildMarkdown è autocontenuta apposta: chrome.scripting ne serializza il
875
+ // sorgente e la esegue nella pagina, dove non esiste il modulo.
876
+ const res = await chrome.scripting.executeScript({
877
+ target: scriptTarget(tabId, frame_id),
878
+ func: buildMarkdown,
879
+ args: [null],
880
+ world: 'MAIN',
881
+ });
882
+ return res?.[0]?.result ?? '';
883
+ }
884
+
856
885
  if (mode === 'accessibility') {
857
886
  // Ritorna una struttura semplificata dell'a11y tree
858
887
  const results = await chrome.scripting.executeScript({
@@ -1639,7 +1668,7 @@ async function cmdFillForm({ fields, submit_selector, tab_id, frame_id }) {
1639
1668
 
1640
1669
  // --- viewport_resize ---
1641
1670
 
1642
- async function cmdViewportResize({ preset, width, height, tab_id }) {
1671
+ async function cmdViewportResize({ preset, width, height, read_only = false, tab_id }) {
1643
1672
  const tabId = await resolveTabId(tab_id);
1644
1673
  const tab = await chrome.tabs.get(tabId);
1645
1674
 
@@ -1656,10 +1685,13 @@ async function cmdViewportResize({ preset, width, height, tab_id }) {
1656
1685
  if (targetW) updateOpts.width = targetW;
1657
1686
  if (targetH) updateOpts.height = targetH;
1658
1687
 
1659
- if (Object.keys(updateOpts).length === 0) throw new Error('Provide preset, width, or height');
1660
-
1661
- await chrome.windows.update(tab.windowId, updateOpts);
1662
- await new Promise((r) => setTimeout(r, 200));
1688
+ if (!read_only) {
1689
+ if (Object.keys(updateOpts).length === 0) throw new Error('Provide preset, width, or height');
1690
+ await chrome.windows.update(tab.windowId, updateOpts);
1691
+ // La finestra si ridisegna in modo asincrono: senza la pausa si leggono le
1692
+ // dimensioni vecchie e si riporta un risultato falso.
1693
+ await new Promise((r) => setTimeout(r, 200));
1694
+ }
1663
1695
 
1664
1696
  const results = await chrome.scripting.executeScript({
1665
1697
  target: { tabId },
@@ -1671,6 +1703,7 @@ async function cmdViewportResize({ preset, width, height, tab_id }) {
1671
1703
  world: 'MAIN',
1672
1704
  });
1673
1705
  const actual = results?.[0]?.result ?? {};
1706
+ if (read_only) return { actual };
1674
1707
  return { requested: { width: targetW, height: targetH, preset: preset || null }, actual };
1675
1708
  }
1676
1709
 
@@ -3508,6 +3541,51 @@ async function cmdHttpRequest({ url, method = 'GET', headers, body, binary = fal
3508
3541
  return { ...common, body: text, size: text.length };
3509
3542
  }
3510
3543
 
3544
+ // --- wait_for_text ---
3545
+
3546
+ /**
3547
+ * Attende che un testo compaia. Si poteva già fare con wait_for_function, ma
3548
+ * quella strada richiede il toggle "Allow user scripts" e un'espressione scritta
3549
+ * a mano: questa no.
3550
+ */
3551
+ async function cmdWaitForText({ text, selector, timeout = 10000, interval = 200, tab_id, frame_id }) {
3552
+ if (!text) throw new Error('Missing required parameter: text');
3553
+ const tabId = await resolveTabId(tab_id);
3554
+
3555
+ const results = await chrome.scripting.executeScript({
3556
+ target: scriptTarget(tabId, frame_id),
3557
+ func: (needle, scopeSel, tout, intv) => new Promise((resolve) => {
3558
+ const started = Date.now();
3559
+ const wanted = String(needle).toLowerCase();
3560
+ const check = () => {
3561
+ const root = scopeSel ? document.querySelector(scopeSel) : document.body;
3562
+ const haystack = (root?.innerText || '').toLowerCase();
3563
+ const at = haystack.indexOf(wanted);
3564
+ if (at !== -1) {
3565
+ resolve({
3566
+ found: true,
3567
+ waited_ms: Date.now() - started,
3568
+ // Contesto attorno alla prima occorrenza: conferma di aver trovato
3569
+ // la cosa giusta senza restituire l'intera pagina.
3570
+ context: (root?.innerText || '').slice(Math.max(0, at - 60), at + wanted.length + 60).trim(),
3571
+ });
3572
+ return;
3573
+ }
3574
+ if (Date.now() - started >= tout) {
3575
+ resolve({ found: false, error: `Text not found within ${tout}ms: ${needle}` });
3576
+ return;
3577
+ }
3578
+ setTimeout(check, intv);
3579
+ };
3580
+ check();
3581
+ }),
3582
+ args: [text, selector || null, timeout, Math.max(50, interval)],
3583
+ world: 'MAIN',
3584
+ });
3585
+
3586
+ return results?.[0]?.result ?? { found: false, error: 'No result' };
3587
+ }
3588
+
3511
3589
  async function cmdSavePage({ tab_id }) {
3512
3590
  const tabId = await resolveTabId(tab_id);
3513
3591
  const blob = await new Promise((resolve, reject) => {
package/package.json CHANGED
@@ -1,7 +1,7 @@
1
1
  {
2
2
  "name": "chrome-bridge-mcp",
3
3
  "mcpName": "io.github.frsorrentino/chrome-bridge",
4
- "version": "1.10.1",
4
+ "version": "1.11.1",
5
5
  "type": "module",
6
6
  "description": "MCP server with 60 tools for browser automation via Chrome extension. Works on ChromeOS/Crostini.",
7
7
  "main": "server/index.js",
@@ -43,6 +43,7 @@ export const MessageType = Object.freeze({
43
43
  UPLOAD_FILE: 'upload_file',
44
44
  WAIT_FOR_NAVIGATION: 'wait_for_navigation',
45
45
  WAIT_FOR_NETWORK_IDLE: 'wait_for_network_idle',
46
+ WAIT_FOR_TEXT: 'wait_for_text',
46
47
  HANDLE_DIALOGS: 'handle_dialogs',
47
48
  FIND_TEXT: 'find_text',
48
49
  NETWORK_RULES: 'network_rules',
@@ -81,7 +82,7 @@ export const MessageType = Object.freeze({
81
82
  });
82
83
 
83
84
  // Versione
84
- export const VERSION = '1.10.1';
85
+ export const VERSION = '1.11.1';
85
86
 
86
87
  // Configurazione
87
88
  export const DEFAULT_PORT = 8765;
@@ -127,6 +128,7 @@ export function getTimeout(type) {
127
128
  if (type === MessageType.WAIT_FOR_ELEMENT
128
129
  || type === MessageType.WAIT_FOR_NAVIGATION
129
130
  || type === MessageType.WAIT_FOR_NETWORK_IDLE
131
+ || type === MessageType.WAIT_FOR_TEXT
130
132
  || type === MessageType.UPLOAD_FILE
131
133
  || type === MessageType.MANAGE_DOWNLOADS
132
134
  || type === MessageType.SAVE_PAGE
package/server/tools.js CHANGED
@@ -187,6 +187,18 @@ export const TOOL_CAPS = {
187
187
 
188
188
  // Parametri ubiqui: un solo testo, così `tab_id` non significa una cosa in un
189
189
  // tool e un'altra nel gemello. test/unit/tool-parameters.test.js lo verifica.
190
+ // save_to: il payload va su disco e nel contesto resta solo il percorso più un
191
+ // sommario. Deliberatamente opt-in per chiamata e non automatico dopo ogni
192
+ // azione: scrivere file per chi non li leggerà è una tassa, e il server non
193
+ // controlla la directory di lavoro del client.
194
+ const saveToField = (what) => z.string().optional()
195
+ .describe(`Absolute path: write ${what} there and return the path instead of the content`);
196
+
197
+ async function savedSummary(path, bytes, extra = {}) {
198
+ await writeFile(path, bytes);
199
+ return { content: [{ type: 'text', text: jsonText({ saved: path, bytes: bytes.length, ...extra }) }] };
200
+ }
201
+
190
202
  const selectorField = (extra = '') => z.string()
191
203
  .describe(('CSS selector; ">>>" pierces shadow DOM. ' + extra).trim());
192
204
  const waitAfter = z.enum(['none', 'navigation', 'networkidle']).optional().default('none')
@@ -509,9 +521,12 @@ export function registerTools(server, wsManager, caps = 'all') {
509
521
  + 'Downscaled to ≤1568px, so fine print may not survive.',
510
522
  {
511
523
  tab_id: tabId,
524
+ save_to: saveToField('the PNG'),
512
525
  },
513
- async ({ tab_id }) => {
526
+ async ({ tab_id, save_to }) => {
514
527
  const data = await send(MessageType.SCREENSHOT, { tab_id });
528
+ const b64 = data?.image ?? data?.data;
529
+ if (save_to && b64) return savedSummary(save_to, Buffer.from(b64, 'base64'), { mimeType: 'image/png' });
515
530
  // data.image è base64 PNG
516
531
  if (data && data.image) {
517
532
  return {
@@ -562,14 +577,16 @@ export function registerTools(server, wsManager, caps = 'all') {
562
577
  selector: z.string().optional().describe('CSS selector; ">>>" pierces shadow DOM. Ignored when ref is given'),
563
578
  ref: z.string().optional().describe('From get_interactives, e.g. "n3"'),
564
579
  force: z.boolean().optional().default(false).describe('Click even if occluded'),
580
+ button: z.enum(['left', 'right']).optional().default('left').describe('right opens the page context menu instead of activating the element'),
581
+ count: z.number().optional().default(1).describe('2 emits dblclick after the two clicks, which is what selects a word or opens an editor'),
565
582
  wait_after: waitAfter,
566
583
  tab_id: tabId,
567
584
  frame_id: frameId,
568
585
  },
569
- async ({ selector, ref, force, wait_after, tab_id, frame_id }) => {
586
+ async ({ selector, ref, force, button, count, wait_after, tab_id, frame_id }) => {
570
587
  const target = resolveTarget(selector, ref, tab_id);
571
588
  const before = await tabSnapshot(tab_id);
572
- const data = await send(MessageType.CLICK, { selector: target, force, frame_id, tab_id });
589
+ const data = await send(MessageType.CLICK, { selector: target, force, button: button ?? 'left', count: count ?? 1, frame_id, tab_id });
573
590
  // Niente attesa se il click non è andato a buon fine (es. elemento occluso)
574
591
  const waited = data?.occluded ? null : await applyWaitAfter(send, wait_after, tab_id);
575
592
  const changed = data?.occluded ? null : pageDelta(before, await tabSnapshot(tab_id));
@@ -617,21 +634,26 @@ export function registerTools(server, wsManager, caps = 'all') {
617
634
  // --- read_page ---
618
635
  server.tool(
619
636
  'read_page',
620
- 'Read the page as text (default), raw HTML, or accessibility tree. Read-only. Expensive on large pages: '
637
+ 'Read the page as text (default), markdown, raw HTML, or accessibility tree. Read-only. markdown keeps '
638
+ + 'headings, links and tables at a fraction of the HTML cost, and is usually the right middle ground. '
639
+ + 'Expensive on large pages: '
621
640
  + 'HTML on a big table costs tens of thousands of tokens for data you then filter anyway — prefer '
622
641
  + 'extract_table or extract for tabular and repeated content, and get_interactives to find click targets.',
623
642
  {
624
- mode: z.enum(['text', 'html', 'accessibility']).default('text').describe('text strips markup, html keeps it (costly), accessibility returns the a11y tree'),
643
+ mode: z.enum(['text', 'markdown', 'html', 'accessibility']).default('text').describe('text strips markup, markdown keeps headings/links/tables far cheaper than html, accessibility returns the a11y tree'),
625
644
  tab_id: tabId,
626
645
  frame_id: frameId,
627
646
  max_length: z.number().optional().default(50000).describe('Max output chars'),
647
+ save_to: saveToField('the page'),
628
648
  },
629
- async ({ mode, tab_id, frame_id, max_length }) => {
649
+ async ({ mode, tab_id, frame_id, max_length, save_to }) => {
630
650
  const data = await send(MessageType.READ_PAGE, { mode, tab_id, frame_id });
651
+ const text = typeof data === 'string' ? data : JSON.stringify(data);
652
+ if (save_to) return savedSummary(save_to, Buffer.from(text, 'utf8'), { mode: mode ?? 'text' });
631
653
  return {
632
654
  content: [{
633
655
  type: 'text',
634
- text: truncateText(typeof data === 'string' ? data : JSON.stringify(data), max_length ?? 50000, 'max_length'),
656
+ text: truncateText(text, max_length ?? 50000, 'max_length'),
635
657
  }],
636
658
  };
637
659
  }
@@ -869,8 +891,9 @@ export function registerTools(server, wsManager, caps = 'all') {
869
891
  + '**returns `found: false` with a reason instead of raising**, so a caller that ignores the result '
870
892
  + 'silently proceeds as if the wait had succeeded. Read-only: waiting changes nothing on the page.',
871
893
  {
872
- condition: z.enum(['element', 'function', 'navigation', 'network_idle']).describe('function needs expression; element needs selector'),
873
- selector: z.string().optional().describe('condition=element'),
894
+ condition: z.enum(['element', 'text', 'function', 'navigation', 'network_idle']).describe('element and text need selector or text; function needs expression'),
895
+ selector: z.string().optional().describe('condition=element; with condition=text it narrows the search to that subtree'),
896
+ text: z.string().optional().describe('Literal text to wait for (condition=text), matched case-insensitively'),
874
897
  expression: z.string().optional().describe('JS expression (condition=function)'),
875
898
  visible: z.boolean().optional().default(false).describe('Element must also be visible'),
876
899
  mode: z.enum(['load', 'spa']).optional().default('load').describe('spa = pushState/popstate/hashchange'),
@@ -880,12 +903,16 @@ export function registerTools(server, wsManager, caps = 'all') {
880
903
  tab_id: tabId,
881
904
  frame_id: frameId,
882
905
  },
883
- async ({ condition, selector, expression, visible, mode, idle_ms, timeout, interval, tab_id, frame_id }) => {
906
+ async ({ condition, selector, text, expression, visible, mode, idle_ms, timeout, interval, tab_id, frame_id }) => {
884
907
  let data;
885
908
  if (condition === 'element') {
886
909
  data = await send(MessageType.WAIT_FOR_ELEMENT, {
887
910
  selector, timeout: timeout ?? 10000, interval: interval ?? 200, visible: visible ?? false, tab_id, frame_id,
888
911
  });
912
+ } else if (condition === 'text') {
913
+ data = await send(MessageType.WAIT_FOR_TEXT, {
914
+ text, selector, timeout: timeout ?? 10000, interval: interval ?? 200, tab_id, frame_id,
915
+ });
889
916
  } else if (condition === 'function') {
890
917
  data = await send(MessageType.WAIT_FOR_FUNCTION, {
891
918
  expression, timeout: timeout ?? 10000, polling_ms: interval ?? 100, tab_id, frame_id,
@@ -997,13 +1024,14 @@ export function registerTools(server, wsManager, caps = 'all') {
997
1024
  + 'corresponding half of the preset, so preset plus width gives a custom width at the preset height. '
998
1025
  + 'A maximized window on ChromeOS ignores the request.',
999
1026
  {
1027
+ action: z.enum(['set', 'get']).optional().default('set').describe('get reports the current viewport without resizing anything'),
1000
1028
  preset: z.enum(['mobile', 'tablet', 'desktop']).optional().describe('375x812, 768x1024, 1440x900'),
1001
1029
  width: z.number().optional().describe('Overrides preset'),
1002
1030
  height: z.number().optional().describe('Overrides preset'),
1003
1031
  tab_id: tabId,
1004
1032
  },
1005
- async ({ preset, width, height, tab_id }) => {
1006
- const data = await send(MessageType.VIEWPORT_RESIZE, { preset, width, height, tab_id });
1033
+ async ({ action, preset, width, height, tab_id }) => {
1034
+ const data = await send(MessageType.VIEWPORT_RESIZE, { preset, width, height, read_only: (action ?? 'set') === 'get', tab_id });
1007
1035
  return {
1008
1036
  content: [{
1009
1037
  type: 'text',
@@ -1877,8 +1905,9 @@ export function registerTools(server, wsManager, caps = 'all') {
1877
1905
  format: z.enum(['lines', 'json']).optional().default('lines').describe('lines is compact; json keeps one object per record'),
1878
1906
  tab_id: tabId,
1879
1907
  max_length: z.number().optional().default(20000).describe('Max output chars'),
1908
+ save_to: saveToField('the records as JSON'),
1880
1909
  },
1881
- async ({ item_selector, fields, max_items, format, tab_id, max_length }) => {
1910
+ async ({ item_selector, fields, max_items, format, tab_id, max_length , save_to }) => {
1882
1911
  max_length = max_length ?? DEFAULT_MAX_OUTPUT;
1883
1912
  const html = await send(MessageType.READ_PAGE, { mode: 'html', tab_id });
1884
1913
  if (typeof html !== 'string') throw new Error('Could not read page HTML');
@@ -1895,6 +1924,7 @@ export function registerTools(server, wsManager, caps = 'all') {
1895
1924
  }
1896
1925
  return row;
1897
1926
  });
1927
+ if (save_to) return savedSummary(save_to, Buffer.from(JSON.stringify({ total: nodes.length, shown: items.length, items }), 'utf8'), { total: nodes.length, shown: items.length });
1898
1928
  if ((format ?? 'lines') === 'json') {
1899
1929
  return { content: [{ type: 'text', text: truncateText(JSON.stringify({ total: nodes.length, shown: items.length, items }), max_length, 'max_items or max_length') }] };
1900
1930
  }
@@ -0,0 +1,197 @@
1
+ #!/usr/bin/env node
2
+ /**
3
+ * Carica e pubblica lo zip dell'estensione sul Chrome Web Store via API.
4
+ *
5
+ * Esiste per togliere di mezzo il TOTP: la dev console non è automatizzabile
6
+ * (la gallery non è scriptabile da un'estensione, e via CDP serve comunque il
7
+ * login Google completo), mentre l'API vuole solo un refresh token revocabile.
8
+ *
9
+ * Credenziali in ~/.config/chrome-bridge/cws.json (chmod 600), MAI nel repo:
10
+ * { "client_id": "...", "client_secret": "...", "refresh_token": "...", "item_id": "..." }
11
+ * Sovrascrivibili da CWS_CLIENT_ID / CWS_CLIENT_SECRET / CWS_REFRESH_TOKEN / CWS_ITEM_ID.
12
+ *
13
+ * Uso:
14
+ * node tools/cws-upload.mjs dist/chrome-bridge-extension-1.11.0.zip # carica come bozza
15
+ * node tools/cws-upload.mjs dist/....zip --publish # carica e pubblica
16
+ * node tools/cws-upload.mjs --auth [porta] # URL di consenso + cattura del codice sul loopback
17
+ * node tools/cws-upload.mjs --exchange <code> [--redirect <uri>] # code → refresh token
18
+ */
19
+ import { createServer } from 'node:http';
20
+ import { readFile, writeFile, mkdir, stat } from 'node:fs/promises';
21
+ import { homedir } from 'node:os';
22
+ import { join, dirname } from 'node:path';
23
+
24
+ const CONFIG = join(homedir(), '.config', 'chrome-bridge', 'cws.json');
25
+ const SCOPE = 'https://www.googleapis.com/auth/chromewebstore';
26
+ // Google ha dismesso il redirect fuori banda (urn:ietf:wg:oauth:2.0:oob): un
27
+ // client nuovo che lo usa riceve "Errore 400: invalid_request". Per i client
28
+ // desktop la strada supportata è il loopback su una porta qualsiasi, che qui
29
+ // viene aperta al volo e chiusa appena arriva il codice.
30
+ function buildAuthUrl(clientId, redirectUri) {
31
+ return 'https://accounts.google.com/o/oauth2/auth?' + new URLSearchParams({
32
+ client_id: clientId,
33
+ redirect_uri: redirectUri,
34
+ response_type: 'code',
35
+ scope: SCOPE,
36
+ access_type: 'offline',
37
+ prompt: 'consent',
38
+ });
39
+ }
40
+
41
+ // Attende un solo codice sul loopback e poi chiude: nessun server che resta su.
42
+ function waitForCode(port) {
43
+ return new Promise((resolve, reject) => {
44
+ const server = createServer((req, res) => {
45
+ const url = new URL(req.url, `http://127.0.0.1:${port}`);
46
+ const code = url.searchParams.get('code');
47
+ const err = url.searchParams.get('error');
48
+ res.writeHead(200, { 'Content-Type': 'text/html; charset=utf-8' });
49
+ res.end(`<h2>${code ? 'Autorizzato. Puoi chiudere questa scheda.' : 'Errore: ' + (err || 'nessun codice')}</h2>`);
50
+ server.close();
51
+ if (code) resolve(code); else reject(new Error(err || 'nessun codice nella risposta'));
52
+ });
53
+ server.on('error', reject);
54
+ server.listen(port, '127.0.0.1');
55
+ setTimeout(() => { server.close(); reject(new Error('nessuna autorizzazione entro 5 minuti')); }, 300000);
56
+ });
57
+ }
58
+
59
+ export async function loadConfig(env = process.env) {
60
+ let file = {};
61
+ try { file = JSON.parse(await readFile(CONFIG, 'utf8')); } catch { /* prima esecuzione */ }
62
+ return {
63
+ client_id: env.CWS_CLIENT_ID || file.client_id,
64
+ client_secret: env.CWS_CLIENT_SECRET || file.client_secret,
65
+ refresh_token: env.CWS_REFRESH_TOKEN || file.refresh_token,
66
+ item_id: env.CWS_ITEM_ID || file.item_id,
67
+ };
68
+ }
69
+
70
+ export function requireFields(cfg, fields) {
71
+ const missing = fields.filter((f) => !cfg[f]);
72
+ if (missing.length) {
73
+ throw new Error(`Manca ${missing.join(', ')} — mettili in ${CONFIG} o nelle variabili CWS_*`);
74
+ }
75
+ }
76
+
77
+ /**
78
+ * Il Web Store risponde 200 anche quando il caricamento è fallito: l'esito vero
79
+ * è in uploadState, e itemError porta il motivo. Trattare 200 come successo è
80
+ * il modo di credere di aver pubblicato senza averlo fatto.
81
+ */
82
+ export function interpretUpload(body) {
83
+ const state = body?.uploadState;
84
+ if (state === 'SUCCESS') return { ok: true, state };
85
+ const reasons = (body?.itemError || []).map((e) => e.error_detail || e.errorDetail || JSON.stringify(e));
86
+ return { ok: false, state: state ?? 'UNKNOWN', reasons };
87
+ }
88
+
89
+ export function interpretPublish(body) {
90
+ const status = body?.status || [];
91
+ const ok = status.includes('OK') || status.includes('PUBLISHED_WITH_FRICTION_WARNING');
92
+ return { ok, status, detail: body?.statusDetail || [] };
93
+ }
94
+
95
+ async function exchange(cfg, code, redirectUri) {
96
+ const res = await fetch('https://oauth2.googleapis.com/token', {
97
+ method: 'POST',
98
+ headers: { 'Content-Type': 'application/x-www-form-urlencoded' },
99
+ body: new URLSearchParams({
100
+ client_id: cfg.client_id, client_secret: cfg.client_secret,
101
+ code, grant_type: 'authorization_code', redirect_uri: redirectUri,
102
+ }),
103
+ });
104
+ const body = await res.json();
105
+ if (!body.refresh_token) throw new Error(`Nessun refresh_token: ${JSON.stringify(body)}`);
106
+ return body.refresh_token;
107
+ }
108
+
109
+ async function saveRefreshToken(cfg, refresh_token) {
110
+ await mkdir(dirname(CONFIG), { recursive: true });
111
+ await writeFile(CONFIG, JSON.stringify({ ...cfg, refresh_token }, null, 2), { mode: 0o600 });
112
+ console.log(`refresh_token salvato in ${CONFIG} (chmod 600)`);
113
+ }
114
+
115
+ async function accessToken(cfg) {
116
+ const res = await fetch('https://oauth2.googleapis.com/token', {
117
+ method: 'POST',
118
+ headers: { 'Content-Type': 'application/x-www-form-urlencoded' },
119
+ body: new URLSearchParams({
120
+ client_id: cfg.client_id,
121
+ client_secret: cfg.client_secret,
122
+ refresh_token: cfg.refresh_token,
123
+ grant_type: 'refresh_token',
124
+ }),
125
+ });
126
+ const body = await res.json();
127
+ if (!res.ok || !body.access_token) {
128
+ // invalid_grant qui significa quasi sempre consent screen in "Testing":
129
+ // in quello stato Google scade i refresh token dopo 7 giorni.
130
+ throw new Error(`Token refresh fallito (${res.status}): ${JSON.stringify(body)}`);
131
+ }
132
+ return body.access_token;
133
+ }
134
+
135
+ async function main() {
136
+ const argv = process.argv.slice(2);
137
+ const cfg = await loadConfig();
138
+
139
+ if (argv.includes('--auth')) {
140
+ requireFields(cfg, ['client_id', 'client_secret']);
141
+ const portArg = argv[argv.indexOf('--auth') + 1];
142
+ const port = Number(portArg) > 0 ? Number(portArg) : 8899;
143
+ const redirect = `http://localhost:${port}`;
144
+ console.log(buildAuthUrl(cfg.client_id, redirect));
145
+ console.error(`in attesa del codice su ${redirect} …`);
146
+
147
+ const code = await waitForCode(port);
148
+ const token = await exchange(cfg, code, redirect);
149
+ await saveRefreshToken(cfg, token);
150
+ return;
151
+ }
152
+
153
+ const exIdx = argv.indexOf('--exchange');
154
+ if (exIdx !== -1) {
155
+ requireFields(cfg, ['client_id', 'client_secret']);
156
+ const code = argv[exIdx + 1];
157
+ if (!code) throw new Error('Uso: --exchange <codice> [--redirect <uri>]');
158
+ const rIdx = argv.indexOf('--redirect');
159
+ const redirect = rIdx !== -1 ? argv[rIdx + 1] : 'http://localhost:8899';
160
+ await saveRefreshToken(cfg, await exchange(cfg, code, redirect));
161
+ return;
162
+ }
163
+
164
+ const zip = argv.find((a) => a.endsWith('.zip'));
165
+ if (!zip) throw new Error('Indica lo zip da caricare, o usa --auth / --exchange');
166
+ requireFields(cfg, ['client_id', 'client_secret', 'refresh_token', 'item_id']);
167
+
168
+ const size = (await stat(zip)).size;
169
+ const token = await accessToken(cfg);
170
+
171
+ const up = await fetch(`https://www.googleapis.com/upload/chromewebstore/v1.1/items/${cfg.item_id}`, {
172
+ method: 'PUT',
173
+ headers: { Authorization: `Bearer ${token}`, 'x-goog-api-version': '2' },
174
+ body: await readFile(zip),
175
+ });
176
+ const upBody = await up.json();
177
+ const upRes = interpretUpload(upBody);
178
+ console.log(`upload ${zip} (${size} B): ${upRes.ok ? 'SUCCESS' : `FALLITO [${upRes.state}] ${upRes.reasons.join(' | ')}`}`);
179
+ if (!upRes.ok) process.exit(1);
180
+
181
+ if (!argv.includes('--publish')) {
182
+ console.log('bozza caricata; ripassa con --publish per mandarla in review');
183
+ return;
184
+ }
185
+
186
+ const pub = await fetch(`https://www.googleapis.com/chromewebstore/v1.1/items/${cfg.item_id}/publish`, {
187
+ method: 'POST',
188
+ headers: { Authorization: `Bearer ${token}`, 'x-goog-api-version': '2', 'Content-Length': '0' },
189
+ });
190
+ const pubRes = interpretPublish(await pub.json());
191
+ console.log(`publish: ${pubRes.ok ? 'OK' : 'FALLITO'} ${pubRes.status.join(',')} ${pubRes.detail.join(' | ')}`);
192
+ if (!pubRes.ok) process.exit(1);
193
+ }
194
+
195
+ if (import.meta.url === `file://${process.argv[1]}`) {
196
+ main().catch((e) => { console.error(String(e.message || e)); process.exit(1); });
197
+ }