@raultov/browser-control-mcp 0.6.0-beta.9 → 0.7.0-beta.11

package/CHANGELOG.md

@@ -10,6 +10,121 @@ from a copy of its code: entries describe the differences from that release.
 
 ## [Unreleased]
 
+## [0.7.0-beta.11] - 2026-10-31
+
+Public **BETA**. Closes gaps with Playwright MCP and chrome-devtools-mcp and
+fixes editors found in real pages: a new `drag` tool (sortable lists,
+boards, sliders, HTML5 drag and drop), `navigate` back and forward in the
+tab's history, `fill_input` with `append`, and `capture_screenshot` of one
+element; `fill_input` now types into Monaco, CodeMirror 6 on Firefox and
+TinyMCE on Firefox, and `snapshot` shows the value of ARIA comboboxes,
+spinbuttons and sliders. Behaviour changes: the default tool list has one
+more tool (about 390 more tokens), `navigate`'s `url` is optional (pass it
+or `history`), and `fill_input` on an editor whose content the page does
+not expose answers that the text was not checked.
+
+### Added
+
+- `drag` tool: drags an element onto another with the mouse. It waits for
+  running animations to end (a sortable list animating its last
+  reordering starts no drag), presses on `from` (checked like
+  `click_element`), moves in 10 steps 40 ms apart to the centre of `to` and
+  releases, so pointer-driven libraries (sortable lists, boards, sliders)
+  and HTML5 drag and drop both work. Chrome does both natively; Firefox's
+  WebDriver input starts a native HTML5 drag but never drops it, leaving the
+  page stuck mid-drag, so there the native `dragstart` is cancelled and the
+  drag replayed by script along the path (the reply says so). Checked on
+  SortableJS and the-internet's drag and drop on both browsers. (#30)
+- `fill_input` takes `append: true`: the text is typed at the end of what
+  the field already holds instead of replacing it, in text fields,
+  textareas and contenteditable elements, rich text editors included (the
+  caret goes to the end of their last text, so the editor records the
+  edit). A number, email or date field, which has no caret to move, is
+  emptied and retyped whole. The reply says `appended`; a field that did
+  not change is an error. Adding to a draft took one `press_key` per
+  character before. (#32)
+- `navigate` takes `history: "back"` or `"forward"` instead of `url`: it
+  moves one entry in the tab's history and waits for that page to load,
+  answering like a URL navigation (`Went back to <url>.`). A page restored
+  from the back-forward cache, or an entry of the same document
+  (`pushState`), answers at once; at either end of the history it fails
+  without moving. Before, going back took `evaluate_js` with
+  `history.back()`, which answered before the page loaded. (#31)
+- `capture_screenshot` takes a `selector` and captures only that element
+  (CSS, `host >>> inner` into shadow roots and iframes, or a snapshot ref):
+  scrolled into view, clipped to its box, whole even when taller than the
+  viewport (at most 16 384 px per side, with a note). Checking one component
+  no longer costs a whole viewport image. (#33)
+
+### Fixed
+
+- On Firefox, `fill_input`, `fill_form` and `press_key` type into an
+  element inside an iframe (TinyMCE): its script focus left the top
+  document's focus on its `<body>`, where Firefox sent the keys, so nothing
+  reached the editor and `fill_input` reported that the field held
+  "something other than the text". Each iframe the element is in is now
+  focused first. (#35)
+- `fill_input` replaces the content of CodeMirror 6 on Firefox: the text
+  was typed before the old content, because CodeMirror puts its own caret
+  back once focused and drops the selection #23 made with a DOM range. A
+  contenteditable element now also gets the keys a user presses after the
+  range: Control+A (Command+A on macOS) to replace, Control+End with
+  `append`. Checked with CodeMirror 6, ProseMirror, Lexical and the test
+  page's editor on both browsers. (#34)
+- `fill_input` types into editors of their own, such as Monaco (VS Code
+  for the web): the snapshot named Monaco's input `textbox "JavaScript
+  Input"`, but `fill_input` refused it (not an input, textarea or
+  contenteditable element) on Chrome, where it is an `EditContext`
+  element, and on Firefox, where it is a hidden textarea, reported success
+  while the editor never got the text. Such an editor (an `EditContext`,
+  a focusable `textbox` role, or a text field without a visible box) now
+  gets a click on its visible container when its input has none, Control+A
+  (Command+A on macOS) or, with `append`, Control+End, then the text; its
+  content cannot be read back, so the reply says it was not checked. (#28)
+- `snapshot` and `find_in_snapshot` show the value of ARIA widgets: an
+  `<input role="combobox">` its text (MUI Autocomplete after a pick), a
+  spinbutton or slider its `aria-valuetext` or `aria-valuenow` (MUI X
+  DatePicker sections), and a button-like combobox the option it shows
+  (Radix Select). Before, only text fields, sliders and `<select>` had one,
+  so the agent could not see that a pick took. (#29)
+
+## [0.6.0-beta.10] - 2026-10-31
+
+Public **BETA**. Fixes four input and snapshot bugs found while driving a
+real web app (Discord): `fill_input` now types into rich text editors
+(Slate, Lexical, ProseMirror) so the editor records the edit, styled
+checkboxes covered by their own label are clicked, controls shown only on
+hover are revealed and clicked, and `snapshot` lists the links and buttons
+inside a clickable container. Behaviour change: a covered element now gets
+a mouse move before the click is refused.
+
+### Fixed
+
+- `click_element`, `hover` and `fill_form` reach controls shown only while
+  the mouse is over something (a message's actions, a row's toolbar): a
+  covered element now gets the mouse first and is checked again 100 ms
+  later. Discord's message "More" button failed with "covered by
+  'div.message…'". An element that stays covered still fails, naming what
+  covers it, without a click. (#26)
+- `snapshot` and `find_in_snapshot` list what a clickable container holds:
+  an element without a role that has a click handler or a pointer cursor
+  was a leaf named from its whole text, so a container that sets
+  `cursor: pointer` hid every link and button inside it (Discord's whole
+  channel list was one line). A clickable that holds interactive elements
+  now lists them; one with text only is still one named line. (#25)
+- `click_element`, `hover` and `fill_form` click a control covered by its
+  own label (or the element its `aria-labelledby` names), as styled
+  checkboxes and switches are: they failed with "covered by 'span…'". An
+  element under anything else still fails, naming it. (#24)
+- `fill_input` and `fill_form` work in rich text editors (Slate, Lexical,
+  ProseMirror, Draft.js): a contenteditable element was emptied with
+  `textContent = ''`, which removed the nodes the editor renders from its
+  own model, so the typed text showed in the page but the editor did not
+  record it (Discord's channel topic was never saved), and the tool
+  reported success. The content is now selected and the first typed key
+  replaces it, as when a user selects all and types; an empty text deletes
+  it through the editing commands. (#23)
+
 ## [0.6.0-beta.9] - 2026-10-31
 
 Public **BETA**. Spends fewer tokens and answers what the agent needs, after

package/README.md

@@ -1,5 +1,7 @@
 # browser-control-mcp
 
+[![Discord](https://img.shields.io/discord/1566088811354656778?label=Discord&logo=discord)](https://discord.gg/JDfcmD73A)
+
 **Browser Control MCP** lets an AI agent control **and debug** real browsers
 at runtime: breakpoints, pause on load, call frames, evaluation in a paused
 frame, stepping and script search on **Chrome (CDP)** and **Firefox
@@ -11,13 +13,13 @@ 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 (0.6.0-beta.9).** This version contains known bugs and differences
+> **Status: BETA (0.7.0-beta.11).** 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
 > in the [issues](https://github.com/raultov/browser-control-mcp/issues).
 
-💬 Questions, ideas or feedback: join us on [Discord](https://discord.gg/TunAHQZ8K).
+💬 Questions, ideas or feedback: join us on [Discord](https://discord.gg/JDfcmD73A), where every release is announced in #announcements.
 
 ## Comparison with Playwright MCP and chrome-devtools-mcp
 
@@ -52,8 +54,8 @@ Tokens are estimated as characters / 4 for every server.
 
 | Task | browser-control-mcp | Playwright MCP | chrome-devtools-mcp |
 | --- | --- | --- | --- |
-| Tool list in context, default tools | 6 040 tokens | 5 330 tokens | 7 770 tokens |
-| Tool list in context, every opt-in tool on | 9 500 tokens | 12 040 tokens | not measured (optional categories) |
+| Tool list in context, default tools | 6 430 tokens | 5 330 tokens | 7 770 tokens |
+| Tool list in context, every opt-in tool on | 9 880 tokens | 12 040 tokens | not measured (optional categories) |
 | Snapshot of Hacker News | **4 588 tokens** | 12 356 tokens | 6 869 tokens |
 | Snapshot of Wikipedia "Web browser" | **12 643 tokens** (cut, searchable) | 34 349 tokens | 27 656 tokens |
 | Snapshot of MDN `<input>` | **12 625 tokens** (cut, searchable) | 158 077 tokens | 76 654 tokens |
@@ -94,7 +96,7 @@ tools are off by default and listed only with `--tool-groups`:
 | `network` | `mock_route`, `list_routes`, `remove_route`, `set_offline` | changes what the page receives |
 
 `--tool-groups all` lists every group. The page, input, observation and
-debugger tools are always listed (33 tools), plus `enable_proxy_auth` with
+debugger tools are always listed (34 tools), plus `enable_proxy_auth` with
 `--proxy-server` and the WebMCP tools with `--enable-webmcp`. A tool of a
 group that is off fails naming the flag. The list is fixed for the server's
 life: it never changes during a session, so clients can cache it.
@@ -116,20 +118,21 @@ see the note.
 | `close_tab` | ✓ | ✓ | |
 | `list_tabs` | ✓ | ✓ | |
 | `switch_tab` | ✓ | ✓ | |
-| `navigate` | ✓ | ✓ | |
+| `navigate` | ✓ | ✓ | `history: "back"` or `"forward"` instead of `url` moves one entry in the tab's history and waits for the page to load (also when it comes from the back-forward cache, which fires no `load`); at either end it fails without moving |
 | `reload` | ✓ | ✓ | |
 | `scroll` | ✓ | ✓ | reports the window positions before and after |
 | `emulate` | ✓ | partial | `emulation` group; `width`/`height`, `device_scale_factor`, `mobile` (touch), `color_scheme`, `reduced_motion`, `forced_colors`, `media` (`print`), `reset`; answers what the page now sees; Firefox: viewport and device pixel ratio only (BiDi has no media or touch emulation yet; those options fail naming them) |
 | `wait_for` | ✓ | ✓ | waits (checking every 100 ms, at most 30 s) until a visible text appears or is gone, a selector is `visible` / `hidden` / `attached` / `detached`, or the URL contains a text; a dialog or debugger pause ends the wait at once |
 | `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) |
-| `snapshot` | ✓ | ✓ | outline of the page: role, name, states, value and a ref (`e12`) per element, through open shadow roots and same-origin iframes; tables without headers (`th`, `thead`, `caption`) are layout and leave only their content, and a cell is named by its text only when it holds nothing else; at most 50 000 characters (`selector` and `depth` narrow it); the same JavaScript on both browsers |
+| `snapshot` | ✓ | ✓ | outline of the page: role, name, states, value and a ref (`e12`) per element, through open shadow roots and same-origin iframes; tables without headers (`th`, `thead`, `caption`) are layout and leave only their content, and a cell is named by its text only when it holds nothing else; comboboxes, spinbuttons and sliders built with ARIA show their value (the field's text, `aria-valuetext` or `aria-valuenow`, the option a button-like combobox shows); an element that is only clickable (a handler or a pointer cursor) is one named line, unless it holds links, buttons or fields, which are listed under it; at most 50 000 characters (`selector` and `depth` narrow it); the same JavaScript on both browsers |
 | `find_in_snapshot` | ✓ | ✓ | `text` (case-insensitive) or `regex` over the whole outline; matching lines under their ancestors, at most 20 |
 | `save_pdf` | ✓ | ✓ | `files` group; prints the tab as the browser's print does (`landscape`, `print_background`, `scale` 0.1–2, `page_ranges` like `1-3, 5`); with `path` (an absolute `.pdf` under the upload roots) writes the file and answers `{path, bytes, pages}`, else returns the PDF as an embedded `application/pdf` resource (at most 5 MB); Chrome `Page.printToPDF`, Firefox `browsingContext.print` |
-| `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; `button` (left/right/middle), `click_count` (2 = double click) and `modifiers`; fails without clicking when it is hidden or covered; `host >>> inner` selectors reach open shadow roots and same-origin iframes (also in `fill_input`, `scroll`, `wait_for`); a dialog or debugger pause the click causes is noted in a successful reply (`{dialog}` or `{paused, location}`) |
-| `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 |
+| `capture_screenshot` | ✓ | ✓ | `format` `png`, `jpeg` (or `jpg`) or `webp`; `full_page` at most 16 384 px per side (a text note says when cut); `selector` captures one element (CSS, `host >>> inner` or a snapshot ref), scrolled into view, whole even when taller than the viewport |
+| `click_element` | ✓ | ✓ | scrolls the element into view; `button` (left/right/middle), `click_count` (2 = double click) and `modifiers`; fails without clicking when it is hidden or covered (its own label on top does not count: styled checkboxes and switches are clicked through it; a covered element first gets the mouse, so controls shown only on hover are revealed and clicked); `host >>> inner` selectors reach open shadow roots and same-origin iframes (also in `fill_input`, `scroll`, `wait_for`); a dialog or debugger pause the click causes is noted in a successful reply (`{dialog}` or `{paused, location}`) |
+| `fill_input` | ✓ | ✓ | picks `<select>` options; fails without typing on elements that cannot take the text; `append: true` types after the content instead of replacing it (a number, email or date field, which has no caret, is retyped whole); an editor of its own (an element with an `EditContext` or a focusable `textbox` role, or the hidden input of a code editor such as Monaco or Ace) gets the keys a user presses: a click on the editor when its input has no visible box, Control+A (Command+A on macOS) or Control+End, then the text, and the reply says its content could not be checked; in a contenteditable element (including rich text editors such as Slate, Lexical, ProseMirror or CodeMirror 6) it selects the content (a range, then Control+A as a user would) and types over it, so the editor records the edit; 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 |
 | `press_key` | ✓ | ✓ | one key (a character, Enter, Tab, Escape, arrows, F1–F12...) with Control / Shift / Alt / Meta held, in `selector` (focused first) or the focused element; a shortcut types nothing |
 | `hover` | ✓ | ✓ | moves the mouse over the element (scrolled into view; fails when hidden or covered) |
+| `drag` | ✓ | ✓ | presses on `from` (checked like `click_element`), moves in steps to the centre of `to` (which must then be in the viewport) and releases: sortable lists, boards, sliders; HTML5 drag and drop is real on Chrome and simulated with script events on Firefox (its WebDriver input starts a native drag but never drops it), and the reply says so |
 | `fill_form` | ✓ | ✓ | several fields in one call: text and `<select>` as `fill_input`, a checkbox or radio takes `true`/`false` and is clicked only when its state differs; stops at the first failing field naming what was done; never echoes values |
 | `upload_file` | ✓ | ✓ | `files` group; sets the files of an `<input type=file>` (Chrome `DOM.setFileInputFiles`, Firefox `input.setFiles`); files must be under the upload roots (`--upload-root`, default the working and temp directories) |
 | `handle_dialog` | ✓ | ✓ | accepts or dismisses the open `alert` / `confirm` / `prompt` / `beforeunload` dialog |
@@ -161,8 +164,8 @@ see the note.
 | `webmcp_get_invocation` | ✓ | ✗ | needs `--enable-webmcp` |
 | `webmcp_list_invocations` | ✓ | ✗ | needs `--enable-webmcp` |
 
-After `click_element`, `hover`, `fill_input`, `fill_form` and `press_key`
-the server waits for the page to settle before answering: until its DOM
+After `click_element`, `hover`, `drag`, `fill_input`, `fill_form` and
+`press_key` the server waits for the page to settle before answering: until its DOM
 stays unchanged for 50 ms (style changes, often animations, aside) and the
 requests the action started finished, at most 1 s, on both browsers. An
 action nothing reacts to costs those 50 ms and keeps its reply; otherwise
@@ -509,7 +512,7 @@ Chrome and Firefox take the same calls:
    with that location.
 5. `evaluate_on_call_frame` reads locals of the paused frame; `step_over`
    moves one statement; `resume` continues. While paused, the input tools
-   (`click_element`, `hover`, `fill_input`, `fill_form`, `press_key`) fail
+   (`click_element`, `hover`, `drag`, `fill_input`, `fill_form`, `press_key`) fail
    at once naming the pause and do nothing: the page would handle their
    events only after `resume`. `evaluate_js` still answers
    while paused, in the page's global scope. A promise it returns cannot

package/npm-shrinkwrap.json

@@ -19,7 +19,7 @@
       "hasInstallScript": true,
       "license": "MIT",
       "name": "@raultov/browser-control-mcp",
-      "version": "0.6.0-beta.9"
+      "version": "0.7.0-beta.11"
     },
     "node_modules/detect-libc": {
       "engines": {
@@ -48,5 +48,5 @@
     }
   },
   "requires": true,
-  "version": "0.6.0-beta.9"
+  "version": "0.7.0-beta.11"
 }

package/package.json

@@ -1,6 +1,6 @@
 {
   "artifactDownloadUrls": [
-    "https://github.com/raultov/browser-control-mcp/releases/download/v0.6.0-beta.9"
+    "https://github.com/raultov/browser-control-mcp/releases/download/v0.7.0-beta.11"
   ],
   "author": "raultov <raultov@protonmail.com>",
   "bin": {
@@ -92,7 +92,7 @@
       "zipExt": ".tar.xz"
     }
   },
-  "version": "0.6.0-beta.9",
+  "version": "0.7.0-beta.11",
   "volta": {
     "node": "18.14.1",
     "npm": "9.5.0"