AnsClaw
中 EN
Download for Windows

Changelog

What each release added and fixed.

56 releases · newest first

v2.1.2.0N 1F 1I 2
New
  • Support batch tag setting for iPhone devices
Fixed
  • Fixed an issue where switching tabs after selecting multiple devices did not take effect correctly
Improved
  • Changed clearing the album to batch processing by directory, reducing waiting
  • Optimized progress and failure prompts when clearing the album across multiple devices
v2.1.1.0A 1
Adjusted
  • The minimum zoom on the device page is restored to 30%
v2.0.0.0N 6F 3I 4A 1
New
  • iPhone (iOS) device support (phase 1): after USB connection, direct screen mirroring, touch, and single-device control
  • Clipboard group control: text copied on the master device can be synced to multiple selected devices
  • Group screenshot: capture screenshots for all selected devices at once and save them separately
  • Album supports filtering by directory (album name)
  • Customizable menu layout: the menu editor can adjust entry order and overflow items
  • Device connection diagnostics: when a connection fails, the cause and troubleshooting direction are given directly
Fixed
  • iOS device touch media authorization failure causes the screen to be visible but unresponsive to taps
  • Sudden frame drops in the main control screen, and jagged edges when zooming
  • The transparent input layer may intercept taps, causing occasional missed taps on the device page
Improved
  • Optimized multi-device screen casting resource scheduling, making simultaneous casting from multiple devices more stable and less resource-intensive
  • Adjusted device card size and fill mode, simplified label interaction
  • Unified management entries such as SMS and contacts into a single menu
  • Narrowed the zoom range on the device page to more practical levels
Adjusted
  • Client-side control settings now follow the range delivered by the server, preventing old configurations from turning off new features
v1.115.1.0N 1F 3
New
  • The client automatically completes iPhone connection preparation and cache image restoration
Fixed
  • After restart, wait for the developer service to be ready before starting screen mirroring
  • Correctly record USB recovery status when the device is offline
  • The deployer only handles the computer environment and dependencies
v1.115.0.0N 2F 2
New
  • The client automatically completes iPhone connection preparation and image restoration
  • Client device-level authorization and background connection re-verification
Fixed
  • The deployer focuses on computer environment and dependency deployment
  • Cancel connection preparation when stopping or switching accounts
v1.114.2.0N 2F 2
New
  • iPhone screen mirroring, control, and device management
  • Device connection preparation and diagnostics entry
Fixed
  • Restore automatic installation of Android device drivers
  • Show a clear prompt when the connection preparation tool version is incompatible
v1.100.0.0N 5F 2I 2A 2
New
  • Group control adds a "Sequential" mode: the master device executes immediately, and the remaining selected devices execute one after another at random intervals
  • Settings page "Group Control Mode" dropdown: Sync / Async / Sequential; the random delay range applies only to Async and Sequential
  • Album management adds "Clear Album": can clear all selected devices, confirmation required before deletion, and if more than 5 devices, the device count must be typed manually
  • Album upload supports uploading to all selected devices at once, with progress displayed per device on separate lines
  • Upload supports interruption: a "Cancel Upload" option is added to the progress card, and files already uploaded remain on the device
Fixed
  • The main control window and the main window title bar cannot be dragged to move the window
  • After dragging a file in, when dragging it out of the window, releasing it at a location where it cannot be dropped, or canceling the drag, the overlay mask does not disappear
Improved
  • Drag-and-drop upload now runs in the background: the small progress card in the bottom-right corner no longer blocks the interface, so you can keep working; leaving the device page no longer interrupts the upload, and a notification bar reports the result when it finishes
  • File management, Photos, App management, Messages, Contacts, Element tools: when multiple devices are selected, only the first device's window opens, with a prompt indicating which device it is
Adjusted
  • The legacy "Asynchronous Group Control" switch is automatically migrated to asynchronous mode and will not revert to synchronous.
  • Removed the "Controlling" badge and the group control mode label from the device card; the master control status is still distinguished by the title bar background color.
v1.99.0.0N 2I 2A 1
New
  • Right-click a device card and select "Report Issue": enter a problem description, attach up to 3 screenshots, and submit them together with the device logs
  • Device logs now include tap diagnostics: they record tap dispatch results, elapsed time, decoding status, and frame fingerprint changes to help pinpoint unresponsive taps
Improved
  • System log slimming: device-related logs such as screen casting service output, group control, and file/media/app operations now go to their respective device logs
  • The admin console can view user-reported exceptions and the accompanying logs
Adjusted
  • ADB trace logging is disabled by default; set the environment variable ANSCLAW_ADB_TRACE=1 to enable output
v1.96.0.0N 2F 1A 2
New
  • Shortcut settings are now split into tabs by scope: Global, Device Page, and Main Control Window, so you can see at a glance which keys apply where
  • Added local shortcuts to the Main Control Window: Select All Text (Ctrl+Alt+A) and Undo (Ctrl+Alt+Z); these can be rebound on the settings page
Fixed
  • The Ctrl+A / Ctrl+Z shortcuts in the main control window conflicted with global shortcuts, so they have been changed to Ctrl+Alt+A / Ctrl+Alt+Z.
Adjusted
  • Ctrl+A, Esc, Ctrl+I, and Ctrl+Shift+T are now global and also work in the main control window
  • Select All and Undo in the main control window now send key combinations directly to the device and no longer depend on whether the input method is ready
v1.95.0.1F 1
Fixed
  • When the device is pressed elsewhere and then moved onto a card and released, it would mistakenly open the main control; now a click is only counted when both the press and the release occur on the card (and no dragging occurs).
v1.95.0.0N 3F 2I 2A 1
New
  • Device page top menu "Album Management": image/video grid, device-side thumbnails, upload (including drag-and-drop), delete, filter, and pagination
  • Logo in the main control window title bar; title shows "Main Control Device-{device name}" and updates in real time when the device is renamed
  • Text input focus gate: when the device does not have input box focus, text or paste is not sent, and the hidden input box is not activated
Fixed
  • Text could still be entered when the device did not have input box focus, and the hidden input box still appeared
  • Album video thumbnails are now generated by frame extraction on the device side; the PC no longer downloads the original video
Improved
  • Album thumbnails are generated on the device side and loaded by page (concurrency 4), so opening a large album no longer sends a large number of requests at once
  • Input focus is cached per device and rechecked after touch/navigation, so continuous input does not check once per character
Adjusted
  • At most 3 album windows can be open at the same time; the album list shows 60 items by default and supports loading more
v1.94.0.0N 7F 4I 3A 2R 1
New
  • Main control window syncs device audio (device audio capture requires Android 13+); a speaker icon toggle is added to the left of the maximize button in the title bar
  • Settings page "Screen-off control": all cast devices turn off screen uniformly; after unlocking or waking, the screen automatically turns off again in about 2 seconds
  • Device page top menu "More functions" (right of lock screen): restart device, shut down device; confirmation before shutdown, manual power-on required after shutdown
  • USB connection automatically enables wireless debugging (adb tcpip 5555); enabled by default in settings, can be disabled
  • Drag files/images onto the device page to upload to the selected device (images go to album, others to Download)
  • Checkbox on the right side of the device card header serves as the selection entry; supports batch tag setting and asynchronous group control with random delay
  • ADB service automatically recovers after being closed externally: device disconnects in about 0.2 seconds, automatically reconnects after adb recovers
Fixed
  • After closing AnsClaw, the device no longer stays with screen off (scrcpy cleanup process restores screen brightness)
  • After device restart or adb restart, casting does not automatically resume and wireless devices do not come back online
  • Closing AnsClaw takes too long (measured reduction from 16.9 seconds to about 2.3 seconds)
  • When wireless transport changes from Detecting to Ready, identity is not resolved, so the device is Ready but no longer appears
Improved
  • When closing the software, cast sessions and IME backend are finalized in parallel, so a single timeout no longer blocks exit
  • Device page box selection changed to toggle semantics; releasing files dragged onto a device card also uploads
  • Device page top "More functions" placed to the right of "Lock screen"
Adjusted
  • Opening/switching main control no longer changes the device selection set; single-click on device card opens main control, double-click removed
  • "Restore auto sort" changed to a settings page switch
Removed
  • Device card selection overlay, main control window settings button
v1.91.0.1N 3F 1I 1
New
  • Device page supports drag-and-drop upload: drag files onto the device page to upload to all checked devices (images go to the album, others to Download)
  • Device page box selection changed to toggle: checked items inside the box are unchecked, unchecked items are checked
  • Usage reporting distinguishes official built-in AI from self-configured models
Fixed
  • Files are not uploaded when dragged onto a device card and released (the page now uses Preview tunnel event handling)
Improved
  • Multi-device uploads run in parallel per device, with files processed sequentially on a single device, global concurrency of 4, and one progress row per device
v1.89.0.1N 4F 1I 2R 1
New
  • Device card selection changed to a checkbox on the right side of the title bar; clicking a card directly opens/switches the main control window
  • Asynchronous group control: the main control executes immediately, other selected devices execute with a random delay (adjustable 0.5–3 seconds)
  • Shortcuts also work when the main control window has focus
  • Main control device card title bar highlighted in the primary color; menu ToolTip shows the current shortcut
Fixed
  • Opening/switching the main control window incorrectly changed the device selection set
Improved
  • Auto-sort changed to a toggle on the settings page; dragging a device card automatically switches to manual and shows a stacked notification
  • UI back unified to the real BACK key (INJECT_KEYCODE)
Removed
  • Ctrl+1…9 shortcuts for selecting devices
v1.86.0.5N 5F 2I 3
New
  • Main window supports drag-and-drop file upload: images go to the album automatically, others to Download, multiple files concurrently
  • Asynchronous group control: the main device executes immediately, other selected devices get random delays, reducing the risk of batch operations in the same second
  • Window snapping: dragging the main window to the left or right side of the main window auto-snaps it, matches height, and adjusts width by aspect ratio
  • Device page supports right-click batch tag setting and one-click restore of automatic sorting
  • Shortcut Ctrl+R opens the device's recent tasks
Fixed
  • Snapped window cannot be dragged out and maximize stops working
  • Drag-and-drop upload overlay flickers on release
Improved
  • Main window removes the minimize button; settings entries are consolidated in the main window settings page
  • Number input fields keep only the input area (up/down adjust and clear buttons hidden)
  • Screen mirroring startup staged waiting, IME startup contention, and UI blocking fixes
v1.76.0.1N 2F 2
New
  • AI navigation experience: successful routes and corrections after going off course are reused across sessions, making navigation more stable when switching pages or apps
  • Automation cards are sorted by creation time
Fixed
  • To-do list out of order after updating a recorded to-do
  • Session context statistics no longer display token numbers
v1.75.0.2N 1F 1
New
  • AI experience memory: control meanings and navigation paths are reused across sessions, saving time on repeated operations
Fixed
  • Session context statistics no longer display token numbers
v1.73.0.0F 1
Fixed
  • Session context statistics no longer display token numbers
v1.72.0.10N 4F 6I 3A 2
New
  • Added a set of shortcuts: select devices, paste computer text into the phone, etc., all via keyboard, with customizable key bindings
  • Added "Element Tool": view which elements are on the current screen and where they are at any time
  • Session list supports multi-select deletion
  • In settings, you can check whether the model in use can call tools properly, to troubleshoot "AI only talks but does nothing"
Fixed
  • Recording was sometimes misjudged as failed and scripts as invalid; now judged truthfully based on actual evidence
  • Some steps in the trial run progress were displayed incompletely (e.g., wait conditions showed only half)
  • Some input fields displayed as white blocks in dark mode
  • When there are many parameter items and the form exceeds the screen, the lower part cannot be fully seen (now scrollable)
  • When the input field does not have focus, text is no longer pasted elsewhere; instead, a clear message indicates the input did not succeed
  • When a local model is configured but no key is filled in, it no longer gets stuck on the configuration page
Improved
  • During recording and playback, screen reading no longer re-fetches each node from the device, significantly reducing wait times
  • Screen reading is prepared in advance at startup, so the first click after starting operations is no longer noticeably laggy
  • The report after recording ends is shorter: it only covers the route, parameters, and uncertain steps
Adjusted
  • Settings page reorganized into categories, making it easier to find a specific item
  • The two internal parameters "result observation time" and "screen reading refresh" for recording have been removed from the settings page
v1.61.1.0N 1F 1
New
  • When the device cannot operate the screen, a clear prompt now indicates which phone setting is missing (for example, Xiaomi requires enabling "USB debugging (Security settings)" and restarting the phone)
Fixed
  • On such devices, AI automation now reports failure truthfully instead of showing a click with no response, and no longer retries repeatedly
v1.61.0.0N 1F 5
New
  • The interface language (Chinese/English) can be switched in the top right corner of the login page
Fixed
  • The email verification code validity period has been extended from 2 minutes to 5 minutes, and the duration stated in the email now matches the actual duration
  • Verification code emails and server prompts now use English for non-Chinese users, so incomprehensible Chinese no longer appears
  • Requesting the verification code again no longer invalidates the code in the previous email
  • When changing the password, a nonexistent account no longer invalidates the verification code, and the original code can be used to retry
  • Users on non-Chinese systems no longer see a Chinese interface when opening the client for the first time
v1.60.2.0F 1
Fixed
  • Improved connection and update experience on overseas networks
v1.60.1.0N 1F 1
New
  • Tap a notification to view its full content
Fixed
  • Fixed an issue where the notification panel background was transparent and hard to see in dark theme
v1.60.0.0N 1
New
  • Added in-app notifications: important messages are shown in the bell in the title bar
v1.59.0.0I 1
Improved
  • Improved the experience of using some features
v1.58.0.0N 4F 4I 4A 1
New
  • Added "App Management": install/uninstall apps, pull installation packages, clear app data, with icons and versions in the list
  • Added "SMS Management": browse, send, delete, clear SMS, search by number or content
  • Added "Contacts Management": browse, add, edit, delete contacts, search by name or number
  • When no model is configured on the AI page, it will guide you step by step to obtain a DeepSeek API Key
Fixed
  • After opening the control window of another device, the "Controlling" indicator on the previous card does not disappear
  • The app list sometimes fails to load (keeps failing)
  • The list does not refresh after uninstalling, deleting, or editing; after opening the window, you have to manually click refresh to display content
  • Some scripts time out and fail right after starting (wait time is now sufficient for cold start)
Improved
  • API keys and other secrets are encrypted and stored locally, no longer in plaintext
  • Text recognition defaults to fast mode for quicker startup; only asks whether to install the acceleration package when a script uses it
  • Large file downloads support resume, continue after interruption without restarting
  • Settings page is simpler: API service is enabled by default, redundant switches removed
Adjusted
  • The script's screen viewing method is split into three modes: "UI tree / text recognition / element recognition", and the script shows which one is used
v1.47.0.0N 2F 3A 1
New
  • Step-by-step onboarding tutorial on first entering the main interface: points out the main areas and common features step by step; can be skipped, and can be rewatched in Settings → Tutorial Learning
  • Let AI help you write skills: discuss clearly what to do, how to verify, and what to do on failure, and it will build the skill directly (a confirmation pops up before writing)
Fixed
  • Clicking the User Agreement and Privacy Policy on the login page may fail to open
  • Starting two clients repeatedly on the same computer would kick each other's login status (prompting that the account is logged in elsewhere); now the second one directly prompts that it is already running
  • The account page sometimes loads very slowly
Adjusted
  • The agreement checkbox on the login page is remembered: once agreed, it will not be required every time; after the agreement content is updated, confirmation will be required again
v1.46.0.0N 3F 2
New
  • Added the User Agreement and Privacy Policy to the login page; you must check the box to agree before logging in, and you can click to view the full text
  • AI can now use image recognition to find targets: when the UI element tree cannot be read, it locates via template or feature matching
  • Automation scripts support template-based locating: they can click the right spot even when no usable UI elements are available
Fixed
  • Fixed the list style in the file management window being inconsistent with other interfaces in the software
  • Fixed a possible error when opening a popup after clicking the agreement link on the login page
v1.45.0.0N 6F 5I 5R 1
New
  • Device file management: browse phone files directly on the device page, download, upload, delete, and display thumbnails for images
  • File management supports operations on multiple devices at once; if a device lacks the file, it is skipped and noted separately
  • Automation scripts support runtime parameters: during recording, extract content, counts, etc. as parameters to fill in at runtime
  • Scripts automatically correct when encountering obstacles and continue running without starting over
  • After a trial run, the script provides a report: which step it reached, how many times it deviated, and whether it succeeded only through corrections
  • Each automatic optimization of a script retains historical versions, allowing rollback to previous versions
Fixed
  • Fixed device auto-sorting failure: new devices always appeared at the bottom
  • Fixed inability to see which step failed when a script run fails
  • Fixed loss of key information during script recording, causing playback to fail at the first step
  • Fixed occasional false failure judgments in automation scripts (a slightly slow interface was judged as failure)
  • Fixed occasional startup failure of the input method service, making SMS and contacts features unavailable
Improved
  • Device list auto-sorting: new devices are placed in ascending order by name, and re-sorted automatically after renaming
  • Scripts can directly open a specified page within an app, with faster cold start
  • Element locating is more tolerant of interface changes; minor UI changes no longer cause failures at every step
  • Input method service starts faster and connects more stably, making SMS and contacts features more reliable
  • Trial run time is significantly reduced
Removed
  • Scripts no longer support Python syntax; old Python scripts need to be re-recorded
v1.31.2.0N 2F 9I 1
New
  • Recorded scripts support "wait a while before continuing", e.g., post a comment every few seconds
  • Recorded scripts are more stable: each step is determined by elements visible on the interface
Fixed
  • Fixed the issue of error exit during script recording
  • Fixed the issue where longer tasks suddenly stop halfway through execution
  • Fixed the issue where script naming suggestions are unrelated to the current task
  • Fixed the issue where the task list is not displayed during script recording
  • Fixed the issue where the thinking timer always shows 0
  • Fixed the issue where Chinese conversations occasionally reply in English
  • Fixed the issue of failing to read the interface when the page loads slowly
  • Fixed the issue where some App pages cannot read interface content
  • Fixed the occasional issue of garbled tags appearing in replies
Improved
  • When script playback fails verification, it will clearly prompt to re-record instead of being considered successful
v1.28.0.0F 1I 2
Fixed
  • Fixed an issue where the interface would freeze and cards would not display fully when there were many devices
Improved
  • Device onboarding is noticeably faster; even plugging in over a hundred devices at once takes only one or two seconds
  • Scrolling on the device page is smoother
v1.27.1.0F 1I 3
Fixed
  • Fixed an issue where screen casting would not start on certain devices
Improved
  • Smoother casting preview when there are many devices
  • Smoother interface when multiple devices are connected
  • More stable input method service when there are many devices
v1.27.0.0F 3R 1
Fixed
  • The AI question panel occasionally became unresponsive or could not be closed; it can now be submitted or canceled normally
  • AI reply language follows you: Chinese if you speak Chinese, English if you speak English; if you specify a language, that language takes precedence
  • AI reads the interface more accurately, with fewer cases of elements not found or wrong positions clicked
Removed
  • Sample skills are no longer preinstalled; you create the skill list yourself
v1.26.3.0I 1
Improved
  • Text input is more reliable: results are verified after input, and if the text was not entered, it retries using the input method.
v1.26.2.0F 1
Fixed
  • Fixed an occasional failure of the input method service to start, which caused SMS, contacts, and other features to be unavailable
v1.26.0.0N 4F 3I 2A 2
New
  • Added global search to the navigation bar: one entry point to search pages, devices, sessions, skills, automations, and scheduled tasks
  • Added wireless connection: scan the local network and connect to devices with wireless debugging enabled in one click
  • Skills now actually take effect: AI loads the content of the skills you enable on demand
  • Support for reading attachments: AI can read the content of text attachments and view images directly
Fixed
  • Fixed the abnormal display of the "Controlling" label on device cards
  • Fixed an issue where a device connected via both USB and wireless appeared as two cards and was counted twice
  • Fixed the main control window always reverting to the default size after opening and not remembering your adjusted size
Improved
  • Greatly expanded the device capabilities callable from scripts (apps, memory, network, screenshot color finding, element tapping, etc.)
  • The main control window now remembers its last position and size
Adjusted
  • Device search is integrated into global search; the device page no longer has a separate search box
  • Removed the "Interrupt Task" button from the device page (it had no actual effect when clicked)
v1.24.3.0F 1I 1A 1
Fixed
  • Fixed an issue where automatic reconnection after network jitter would switch to another account
Improved
  • User management (admin side): changing level, extending expiry, resetting password, and banning/unbanning are merged into a single edit dialog that submits only the changed items
Adjusted
  • Downgrade is not allowed before expiry: the client button is grayed out, and the server also validates this on order placement and callback settlement
v1.24.2.0N 6F 5I 7A 4R 3
New
  • Screen observation merged into a single entry: one call first returns the interface structure (elements/text/coordinates), automatically supplements with text recognition (OCR) when content cannot be read, then falls back to element detection (YOLO)
  • Added "Context Statistics": in the session "⋯" menu, view the composition of what is actually sent to the model in this round (how much is taken by system prompt / loop instructions / tool definitions / historical messages respectively)
  • Added settings menu (gear button) to the input box: centralized entry for approval whitelist, model management, MCP client, MCP server
  • Agent architecture upgraded to a single Harness "think-act" loop + background sub-agents (automated replay / scheduled tasks / Python escape hatch), making long tasks more stable
  • Device selection panel supports single selection: picking one makes it the operation target and automatically opens its control window (human and machine see the same screen)
  • Unified automation scheduling: batch replay executes per device with zero LLM, failures automatically escalate to AI repair and write back to the script (self-learning)
Fixed
  • Fixed status not displaying after tool call completion (green check not appearing, results not visible)
  • Fixed program exception when clipboard is occupied by other programs
  • Fixed error when toggling input method service (detection exception when device-side service is not started)
  • Fixed ADB status query throwing exception when service is not started
  • Fixed task list last step not updating
Improved
  • Prompt system reorganized: rules deduplicated and reduced by about 70%, tool descriptions maintained only in the tool's own description (saves tokens, less deviation)
  • Tool steps and task list status clearer: green check for completion, red cross for failure; task list will not stay in "in progress"
  • Screen casting rendering introduces global frame budget, multiple devices casting simultaneously no longer freezes
  • Device control window linked with main window: always above main window, minimizes/closes with main window, no longer duplicated in taskbar
  • Clipboard read/write more stable: automatically waits and retries when occupied by other programs, no longer throws exceptions
  • Script mode prompts released with client version (no longer dependent on online prompt files), Python runtime built into the package
  • ADB startup consolidated to device page, avoiding connection errors during startup; input method service detection no longer falsely reports exceptions
Adjusted
  • AI operation target unified to "single master device": whichever device's control window is open, AI operates that device (device page checkboxes only for batch casting and automation batches)
  • When AI has not yet determined the target, first pop up device selection panel for user to pick one, then start observation and operation
  • Script mode entry changed to "mode" dropdown selection in input box
  • Sensitive operations (executing device commands, deleting SMS/call logs/contacts, making calls) continue to require approval; automatically allowed when whitelist is hit or unattended
Removed
  • Removed context compression related features and progress display
  • Removed "select multiple at once" and staggered execution mode (to avoid batch operations being out of sync with actual device states)
  • Removed /record, /plan text commands (script mode now entered via mode dropdown)
v1.21.2.1N 3
New
  • Reply language anchor: non-Chinese messages automatically include a language instruction, fixing the drift where English questions received Chinese replies
  • Hardened the language iron rule in Agent prompts (locked throughout + restated at the end)
  • Deployer self-update (v1.0.4) + Python runtime shipped with the package, Feishu remote control available offline
v1.21.1.8N 3F 2
New
  • Model management: built-in vendor catalog only, online retrieval/unified form addition (context presets)
  • ChatInputBox model selection dropdown (clicking when no models opens the model management window)
  • Agent prompt language following upgraded to a hard rule
Fixed
  • Feishu inbound switched to official SDK sidecar (C# raw WS event truncation)
  • Fixed model addition radio-button conflict
v1.19.0.32N 2I 3
New
  • OcrScreen/OcrFile results now return position coordinates (box)
  • AI reply cleaning: strip bare tool-call templates leaked by the model (including full-width ｜ variants)
Improved
  • DumpUi is now the default first choice for UI analysis, with multiple calls and verification at each step
  • CaptureScreen/OcrFile descriptions avoid chaining with OcrScreen
  • TOOLS.md observation section rewritten: DumpUi first, OcrScreen as supplement
v1.19.0.31I 4
Improved
  • System prompt restructured: IDENTITY/SAFETY/TOOLS/AGENTS/CONTEXT aligned with mainstream naming
  • Page state observation priority: Activity (lightweight) → DumpUi → OCR → YOLO (visual advantage)
  • Scheduled task/ScheduleCreate Q&A guidance merged into TOOLS
  • DumpUi outputs all nodes by default
v1.19.0.27F 1I 1
Fixed
  • OCR garbled text: replaced PaddleOCR.dll at runtime with a fixed version (medium model compatible)
Improved
  • DumpUi now outputs all nodes by default
v1.19.0.23N 2F 1
New
  • Collect information with a Q&A control before creating a scheduled task (ScheduleCreate guide prompt, FSS online update)
  • MCP service available only in Pro (CanMcp permission gating: login startup + configuration interception + editable in admin panel)
Fixed
  • Confusing AI script naming suggestions (model echoed prompt instructions): rewrote naming prompt (plain text + examples + output only 3 lines) + parse and filter echoed instructions + fallback task name; prompt moved to Prompts/ for FSS online update
v1.19.0.20F 1
Fixed
  • Confusing AI suggestions for script naming (model echoed prompt instructions) → rewrote the prompt (plain text + examples + output only 3 lines) + filtered echoed instructions during parsing + fallback task name
v1.19.0.19F 2
Fixed
  • Fixed startup crash: CountdownDismissButton default style OverridesDefaultStyle is invalid
  • Heartbeat keep-alive + login reconnect
v1.19.0.18F 2
Fixed
  • First login after idle fails (connection silently dies): fixed at the root with automatic reconnect retries + heartbeat keepalive
  • Fixed brief disconnections during login/calls
v1.19.0.9F 3
Fixed
  • Fixed occasional main window freeze after startup (prompt loading deadlock caused the UI to become unresponsive)
  • Fixed HTTP 400 error when AI sends messages: Tool names must be unique (duplicate tool name conflict)
  • Restored the sub-agent delegation tool (RunSubAgent not registered regression)
v1.18.0.14N 3I 2
New
  • MCP client: connects to external MCP servers (stdio/HTTP), integrating their tools into the AI Agent toolset
  • MCP server: exposes all AnsClaw tools externally (stdio + HTTP, toggle in settings, off by default)
  • Chat UI: tool block expansion shows only IN/OUT, bubbles remove time/name, copy button always visible
Improved
  • AI multi-device selection and synchronized execution (SelectDevices/batch execution/device occupancy mutex)
  • Chinese logs converted to English, IME badge flicker fixed
v1.18.0.11F 3I 1
Fixed
  • IME log double-write: each IME log entry appears twice (Emit writes directly to DLog and the Progress subscription callback writes it again)
  • IME watchdog false restart: when UiAutomation is inactive, reuse priority incorrectly judges a still-live process as alive, without actually force-stopping and rebuilding
  • IME reuse check now includes a real ping: TCP Connected is a stale state and untrustworthy, preventing false reuse
Improved
  • Release rule: version upgrades must keep Version/AssemblyVersion/FileVersion in sync across all three
v1.18.0.10N 1
New
  • Verification: -Changelog compatible path
v1.15.0.0I 1A 1
Improved
  • Client static resources now connect directly to FSS; removed the server-side resource API to reduce server load
Adjusted
  • Local cache now uses .res binary cache (to prevent accidental modification)
v1.13.0.45F 2I 3
Fixed
  • Fixed an error when the TodoWrite tool's title parameter is missing
  • Hardened DetectScreenUi(YOLO) against native crashes
Improved
  • Account-level settings cloud sync: settings.json + log-settings.json stored per account in FSS, automatically synced when switching computers
  • AI system prompts moved online: loaded from FSS for hot updates, and kept confidential without exposing tool names/system prompts
  • Chat input box attachment context injection: file paths are passed to the AI to read on send, and automatically cleared after sending
v1.13.0.38F 4I 3
Fixed
  • Red box after switching the quality dropdown to English
  • Casting baseline tier now calculated by actual casting count (over-limit devices no longer lower image quality)
  • GPU fallback not taking effect when OCR mode is unauthorized
  • API service no longer enabled by default for free users
Improved
  • Client account permissions fully implemented: device limit enforcement / real-time downgrade on expiry / AI monthly quota / automation / scheduled tasks / skills / OpenAPI restrictions
  • Settings page adds manual OCR mode selection (CPU/GPU), effective after login according to settings
  • OCR initialization moved to after login; GPU acceleration available for Max/Pro
v1.13.0.32I 4
Improved
  • AnsLogger unified refactor: unified Log entry (Project auto-derived) + full device log DLog configuration (Show/Hide/whitelist/blacklist/Toast)
  • Log configuration persistence: full log-settings.json mapping; runtime Reconfigure / startup loading / json-change hot reload all use the same configuration
  • LogViewer left-side device/log file list + search box + delete
  • Log notification callbacks: Snackbar/MessageBox/ContentDialog/InfoBar/Flyout + bottom-bar toast (performance debouncing)
v1.13.0.31N 1I 3
New
  • Added casting quality presets on the settings page (Smart Optimization / HD / Semi-HD / SD / Smooth)
Improved
  • Refactored Smart Optimization for casting: content-aware scoring + symmetric up/down switching + quality preset system
  • Real-time FPS and smoothness status labels on device cards
  • Refactored the FSS version mechanism: the versions/ directory is now the source of truth, making releases/upgrades more reliable

Older releases are listed in the Chinese changelog.

Source: product release records. Commercial items such as top-ups and ads are not shown here.