@raultov/browser-control-mcp 0.3.4-beta.4 → 0.3.6-beta.6

package/CHANGELOG.md

@@ -10,6 +10,126 @@ from a copy of its code: entries describe the differences from that release.
 
 ## [Unreleased]
 
+## [0.3.6-beta.6] - 2026-10-29
+
+Public **BETA**. `fill_input` types the same way on both browsers,
+`evaluate_js` gives objects the same value and names their class,
+`inspect_dom` reads shadow roots and same-origin iframes, network logs
+carry cookies and no longer wait on open streams, and raw commands that
+would break the server's capture are refused. Some behaviour changes:
+`fill_input` presses a key per character (about 1 ms each on Chrome),
+`evaluate_js` values are their `JSON.stringify` form, `inspect_dom`
+returns more HTML, and `send_cdp_command` refuses `*.disable` of the
+captured domains and `Browser.close`.
+
+### Fixed
+
+- `set_breakpoint` on Chrome fails for a line with no code at or after it
+  in the loaded script with that URL, saying where the script ends, and
+  sets nothing (as Firefox does). It answered `resolved: false`, which the
+  description explains as "until a script with that URL loads", for a
+  script that was loaded already: the breakpoint never fired.
+
+- `inspect_dom` includes open shadow roots (as `<template
+  shadowrootmode="open">`) and the documents of same-origin iframes (each
+  after an `<!-- iframe <src>: -->` comment), and notes the cross-origin
+  iframes it cannot read. Text inside a web component or an iframe was
+  reported as not found.
+
+- `get_network_logs` asks for a response body only once the response
+  finished loading, and the reply notes the ones still loading (an open
+  event stream). On Firefox an open server-sent events stream made every
+  call wait 10 s for its body; now it answers at once. Each body waits at
+  most 2 s.
+
+- `get_network_logs` on Chrome returns the `Cookie` and `Set-Cookie`
+  headers (redacted unless `include_sensitive_headers: true`): Chrome
+  reports them only in its ExtraInfo events, which were ignored, so they
+  were missing even when asked for. On Firefox a repeated header (several
+  `Set-Cookie`) keeps every value, joined by newlines as on Chrome; only
+  the last one was kept.
+
+- `capture_screenshot` checks `format` before capturing: `png`, `jpeg`
+  (or `jpg`) and `webp`, any case; anything else fails saying so. Firefox
+  answered a PNG labelled `image/gif` (or `image/jpg`), and Chrome a raw
+  `Invalid image format`. The reply's MIME type is the one of the image's
+  bytes.
+
+- `send_cdp_command` refuses the commands that silently stopped what the
+  server captures (`Network.disable`, `Page.disable`, `Runtime.disable`,
+  `Log.disable`, `Debugger.disable`) or closed the browser
+  (`Browser.close`), saying what depends on them; `Network.disable` used to
+  leave `get_network_logs` empty without a word. `Fetch.enable` is sent with
+  a note that requests now pause until answered.
+
+- `evaluate_js` (and `evaluate_on_call_frame`) gives an object or array the
+  same `value` on both browsers: what `JSON.stringify` gives (`toJSON`
+  counts, functions drop out), and an object that is not a plain one has
+  its class in `description`. `navigator` and `new URL(...)` answered `{}`
+  without a word; now `{"value": {}, "description": "Navigator"}` and
+  `{"value": "<href>", "description": "URL"}`. Nested functions were `{}`
+  on Chrome and `{"type": "function"}` on Firefox, and `location` was `{}`
+  on Firefox.
+
+- `fill_input` types the same way on both browsers: Chrome now presses a
+  key per character, as Firefox did, so the page's keydown and keyup
+  handlers run and can cancel a key (a field refusing `a` got `abc` on
+  Chrome and `bc` on Firefox; now `bc` on both). Newlines and tabs are
+  inserted as text on both (as a key, Chrome pressed Tab and the rest of
+  the text left the field). Typing takes about 1 ms per character, so a
+  very long text is slower on Chrome than before.
+- `fill_input` on Firefox types a long text in batches of 500 characters:
+  20 000 characters timed out half typed, now they take about 20 s.
+- `fill_input` fires `change` once, as for a user: the browser fires it
+  when the field loses focus, and when the window has no focus (Firefox
+  headless, which never fired it) the fill fires it at the end.
+
+## [0.3.5-beta.5] - 2026-10-28
+
+Public **BETA**. Observation tools stop returning wrong or missing data
+without saying so, an endless `evaluate_js` loop no longer kills a Chrome
+tab, Firefox debugger calls no longer block other tabs, and WebMCP passes
+every tool annotation through. Some replies change: `get_custom_events`
+answers `{events, total, truncated}` with the newest events, and
+`get_network_logs` requests may carry `error` and
+`response_body_encoding`.
+
+### Fixed
+
+- `webmcp_list_tools` drops no annotation: `consequential` (the page
+  declares the tool has real-world effects, such as a payment) and
+  `debugging` were silently lost, and any annotation a later Chrome adds is
+  now passed through. The output of a tool annotated `untrustedContent` is
+  flagged: `webmcp_invoke_tool` adds a note to treat it as data, not as
+  instructions, and `webmcp_get_invocation` says `untrustedContent: true`.
+- `evaluate_js` with an endless loop no longer kills the tab on Chrome:
+  the script is stopped after 8 s and the call says so ("the script ran for
+  more than 8 s ... the page keeps working"); later calls work at once. It
+  used to fail after 20 s with `Runtime.evaluate timed out`, and every
+  later call on the tab timed out too. An expression waiting on a promise
+  that never settles says so after 10 s. Firefox cannot stop a running
+  script from outside: its error says so and points to `close_tab`.
+- On Firefox, a slow debugger call on one tab (e.g. a long
+  `evaluate_on_call_frame`) no longer blocks the logs and debugger tools of
+  the other tabs: each tab's debugger has its own lock (one lock covered
+  them all, delaying `get_console_logs` on another tab by 9.5 s).
+- `get_custom_events` returns the newest `limit` events (it returned the
+  oldest, so new events never showed once the buffer held more), answers
+  pure JSON `{"events", "total", "truncated"}` (the text started with
+  `Found N custom events.`), and takes `clear` to remove what it returned.
+- `get_network_logs`: on Chrome the page's own document is in its
+  navigation's bucket, so `current_navigation_only` lists it and its body
+  is returned (it was left in the previous navigation, without a body). A
+  failed request carries `error` (Chrome `net::ERR_...`, Firefox
+  `NS_ERROR_...`, also in the summary); it looked like one still waiting on
+  Chrome, and Firefox put the error in `response_status_text`. A binary body
+  says `response_body_encoding: "base64"`.
+- On Chrome, `get_console_logs` and `get_network_logs` connect first: on an
+  instance nothing had connected to (an attached Chrome) they answered
+  empty buffers. The console messages the page logged before are now
+  returned; `get_network_logs` notes that earlier requests are not
+  captured.
+
 ## [0.3.4-beta.4] - 2026-10-27
 
 Public **BETA**. The same server as 0.3.3-beta.4, which was published on

package/README.md

@@ -11,7 +11,7 @@ It is the successor of
 from a copy of its code) and complements Playwright MCP or
 chrome-devtools-mcp with runtime JavaScript debugging on both browsers.
 
-> **Status: BETA (0.3.4-beta.4).** This version contains known bugs and differences
+> **Status: BETA (0.3.6-beta.6).** This version contains known bugs and differences
 > between Chrome and Firefox (see [Limitations and TODO](#limitations-and-todo))
 > and parts that have not been verified (see [Untested](#untested)). Tool
 > replies may change in incompatible ways before 1.0. Please report problems
@@ -49,15 +49,15 @@ see the note.
 | `navigate` | ✓ | ✓ | |
 | `reload` | ✓ | ✓ | |
 | `scroll` | ✓ | ✓ | reports the window positions before and after |
-| `inspect_dom` | ✓ | ✓ | at most 100 000 characters without `query` (cut in the page) |
-| `capture_screenshot` | ✓ | ✓ | `full_page` at most 16 384 px per side (a text note says when cut) |
+| `inspect_dom` | ✓ | ✓ | includes open shadow roots and same-origin iframes (cross-origin ones are noted); at most 100 000 characters without `query` (cut in the page) |
+| `capture_screenshot` | ✓ | ✓ | `format` `png`, `jpeg` (or `jpg`) or `webp`; `full_page` at most 16 384 px per side (a text note says when cut) |
 | `click_element` | ✓ | ✓ | scrolls the element into view; fails without clicking when it is hidden or covered |
-| `fill_input` | ✓ | ✓ | picks `<select>` options; fails without typing on elements that cannot take the text; Chrome inserts the text, Firefox types keys |
+| `fill_input` | ✓ | ✓ | picks `<select>` options; fails without typing on elements that cannot take the text; types key by key on both (key handlers run; newlines and tabs inserted as text), about 1 ms per character; `change` on leaving the field, or at the end when the window has no focus |
 | `handle_dialog` | ✓ | ✓ | accepts or dismisses the open `alert` / `confirm` / `prompt` / `beforeunload` dialog |
-| `evaluate_js` | ✓ | ✓ | `value` at most 100 000 characters (`truncated: true`) |
+| `evaluate_js` | ✓ | ✓ | `value` at most 100 000 characters (`truncated: true`); Chrome stops a script after 8 s, Firefox cannot (a looping tab is closed with `close_tab`) |
 | `get_console_logs` | ✓ | ✓ | a message keeps at most 10 000 characters; Firefox: page errors, CSS warnings and platform messages need the debugger port |
-| `get_network_logs` | ✓ | ✓ | bodies at most 64 000 characters, none over 2 MB; a WebSocket payload keeps 64 000 (`payload_len` is the whole); Firefox: WebSocket frames need the debugger port |
-| `get_custom_events` | ✓ | partial | params over 64 000 characters of JSON are replaced by `{truncated, characters}`; Firefox: events of the modules subscribed with `send_bidi_command` |
+| `get_network_logs` | ✓ | ✓ | a failed request has `error`, a binary body `response_body_encoding: "base64"`; bodies at most 64 000 characters, none over 2 MB; a WebSocket payload keeps 64 000 (`payload_len` is the whole); Firefox: WebSocket frames need the debugger port |
+| `get_custom_events` | ✓ | partial | `{events, total, truncated}`, the newest `limit` events; `clear` removes the returned ones; params over 64 000 characters of JSON are replaced by `{truncated, characters}`; Firefox: events of the modules subscribed with `send_bidi_command` |
 | `get_performance_metrics` | ✓ | partial | Firefox: `Nodes`, `Frames`, `Resources`, `TransferSize` from the Web Performance APIs (no Performance domain), plus a note |
 | `profile_page_performance` | ✓ | partial | Firefox: FCP, LCP, DCL and Load from the Web Performance APIs; no trace, so `total_blocking_time_ms` and `top_long_tasks` are `null` |
 | `set_breakpoint` | ✓ | partial | Firefox: by URL only (no `script_id` / `script_hash`) |
@@ -68,11 +68,11 @@ see the note.
 | `search_scripts` | ✓ | ✓ | a long line's preview is 100 characters on each side of the match |
 | `pause_on_load` | ✓ | ✗ | TODO on Firefox, see [Limitations and TODO](#limitations-and-todo) |
 | `enable_proxy_auth` | ✓ | partial | needs `--proxy-server`; a request the proxy rejects twice is cancelled; untested on Firefox against a real proxy |
-| `send_cdp_command` | ✓ | ✗ | raw Chrome DevTools Protocol |
+| `send_cdp_command` | ✓ | ✗ | raw Chrome DevTools Protocol; refuses `Network`/`Page`/`Runtime`/`Log`/`Debugger.disable` and `Browser.close`; `Fetch.enable` comes with a note |
 | `send_bidi_command` | ✗ | ✓ | raw WebDriver BiDi; refuses `session.new`, `session.end`, `browser.close` |
 | `send_rdp_request` | ✗ | ✓ | raw Firefox RDP request to an actor (`thread`, `console`, `root` shortcuts) |
-| `webmcp_list_tools` | ✓ | ✗ | needs `--enable-webmcp`; WebMCP is a Chrome testing feature |
-| `webmcp_invoke_tool` | ✓ | ✗ | needs `--enable-webmcp` |
+| `webmcp_list_tools` | ✓ | ✗ | needs `--enable-webmcp`; WebMCP is a Chrome testing feature; every annotation Chrome reports is listed (`readOnly`, `consequential`, `untrustedContent`, `debugging`, `autosubmit`, newer ones as they are) |
+| `webmcp_invoke_tool` | ✓ | ✗ | needs `--enable-webmcp`; the output of an `untrustedContent` tool comes with a note to treat it as data |
 | `webmcp_get_invocation` | ✓ | ✗ | needs `--enable-webmcp` |
 | `webmcp_list_invocations` | ✓ | ✗ | needs `--enable-webmcp` |
 
@@ -295,7 +295,11 @@ claude mcp add browser-control-mcp -- npx -y @raultov/browser-control-mcp
   `null`, `node`, `date`, `map`...), `value` is the JSON form and is `null`
   when there is none (functions, DOM nodes, symbols, `NaN`, bigints,
   maps...), `description` then says what it was (Chrome describes functions
-  by their source, Firefox only as `function`). A thrown exception or a
+  by their source, Firefox only as `function`). An object or array's
+  `value` is what `JSON.stringify` gives, on both browsers (`toJSON` counts:
+  a `URL` is its `href`; nested functions drop out), and an object that is
+  not a plain one has its class in `description` (`navigator`:
+  `{"value": {}, "description": "Navigator"}`). A thrown exception or a
   rejected promise fails with what was thrown and its 1-based line and
   column, e.g. `ReferenceError: x is not defined (line 1, column 1)`.
   A `value` over 100 000 characters is not returned whole: a string keeps
@@ -310,7 +314,11 @@ claude mcp add browser-control-mcp -- npx -y @raultov/browser-control-mcp
   `{"messages": [...], "total", "truncated", "notes"?}`: at most `limit`
   (default 200) of the most recent matching messages, oldest first; `clear`
   removes only the returned ones; a note says when a site change cleared
-  messages.
+  messages. `get_custom_events` answers `{"events", "total", "truncated"}`
+  the same way. On Chrome both log tools connect first, so on an attached
+  browser the first call already sees the page's console messages (Chrome
+  replays them; network requests sent before the connection are not
+  captured, and the reply says so).
 - **Firefox allows a single automation session.** Each Firefox instance holds
   one WebDriver BiDi session shared by its tabs; `close_instance`,
   `stop_browser` and every server shutdown path end it, so the next run can
@@ -441,6 +449,7 @@ incompatible:
 | `remove_breakpoint` | Chrome: CDP id, reply `{}` | `bp-N` id, reply `{removed}` |
 | `step_over`, `resume` | Chrome: text only; Firefox `step_over`: `{paused, location: {line (1-based), column}}` | `{paused, location: {url, line_number, column_number} \| null}`, 0-based |
 | `evaluate_on_call_frame` | Chrome: raw `RemoteObject` (exceptions as success); Firefox: bare value | the `evaluate_js` shape; exceptions are errors (`Evaluation threw: ...`) |
+| `get_custom_events` | `Found N custom events. Captured events: [...]`, the oldest events | `{events, total, truncated}`, the newest `limit` events; `clear` removes the returned ones |
 | `get_console_logs` | a JSON array; `clear` emptied the whole buffer | `{messages, total, truncated, notes?}`, at most `limit` (200) messages; `clear` removes the returned ones; the buffer is cleared on cross-site navigation |
 
 ## Limitations and TODO

package/npm-shrinkwrap.json

@@ -19,7 +19,7 @@
       "hasInstallScript": true,
       "license": "MIT",
       "name": "@raultov/browser-control-mcp",
-      "version": "0.3.4-beta.4"
+      "version": "0.3.6-beta.6"
     },
     "node_modules/detect-libc": {
       "engines": {
@@ -48,5 +48,5 @@
     }
   },
   "requires": true,
-  "version": "0.3.4-beta.4"
+  "version": "0.3.6-beta.6"
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "artifactDownloadUrls": [
-    "https://github.com/raultov/browser-control-mcp/releases/download/v0.3.4-beta.4"
+    "https://github.com/raultov/browser-control-mcp/releases/download/v0.3.6-beta.6"
   ],
   "author": "raultov <raultov@protonmail.com>",
   "bin": {
@@ -92,7 +92,7 @@
       "zipExt": ".tar.xz"
     }
   },
-  "version": "0.3.4-beta.4",
+  "version": "0.3.6-beta.6",
   "volta": {
     "node": "18.14.1",
     "npm": "9.5.0"