Improves Session Compaction and MCP Status Display

• Improved retention of recent responses and tool results so they are no longer lost after restarting and resuming a session
• Fixed services pending authentication not correctly displaying their status in the /mcp command interface list

Session Reliability Improvements and Enterprise Mode Controls

• The built-in /verify skill now requires explicit user invocation and can no longer be triggered automatically by the model
• Fixed an error when resuming very large sessions with /resume
• Added support for enterprise admins to disable Auto and YOLO modes
• Fixed Windows installation corrupting Chinese directory names in the user PATH or adding duplicate Qoder CLI paths when using a non-UTF-8 code page
• Improved retry logic so that after a successful retry of an oversized request, a subsequent oversized request can be retried again
• Improved change detection for read files to prevent compaction from being triggered directly when content reaches the limit without any conversation
• Fixed an issue where retrying in the same session could not resume writing conversation history after a file-busy condition had cleared
• Applications using CodeMode can now display individual tool-call statuses in real time and restore tool-call records from session history

Session status reporting via OSC 7501 and reconnection fixes

• Added support for reporting session status to the terminal via the OSC 7501 protocol
• Fixed an issue where recent messages could be missing when resuming a compacted session after a remote reconnection

Introducing CodeMode and More Reliable Session Continuity

• Fixed request failures in long sessions caused by accumulated images exceeding the limit, keeping at most the 50 most recent images per request
• Fixed occasional crashes caused by race conditions when reading Git branch information at session startup
• Shortened the default wait time for MCP initialization in non-interactive sessions from 30 seconds to 5 seconds
• Improved environment context management by moving environment and Git information out of the system prompt and syncing working directory changes promptly
• Improved memory extraction logic
• Improved automatic resumption after network fluctuations and temporary service errors, reducing unexpected task interruptions
• Added the CodeMode tool, supporting multi-step tool orchestration and on-demand invocation of MCP tools to reduce context usage. Enable it with the environment variable QODER_ENABLE_CODE_MODE=1 and select on or only mode in /settings
• Fixed an error when switching an ACP session to the yolo (Bypass Permissions) mode, so ACP clients can now create and run sessions with Bypass Permissions
• You can now retry in the same session after restoring transcript file permissions, without losing the current worktree, transcript path, or file rollback points
• Fixed being unable to start Grep searches on a single file under Windows network shared drives or UNC/device-style paths

Performance improvements and Auto Mode enhancements

• Improved the WebFetch tool to apply different handling based on model rates
• Improved Subagent startup speed so Git operations no longer block startup
• Improved Auto Mode approvals for long sessions and strengthened protection for sensitive configuration writes
• Fixed write operations being rejected in dontAsk mode when using Memory via the SDK on Windows

Stability Fixes and Context Loading Improvements

• Fixed Hook progress from background Agent incorrectly showing in the TUI
• Rules in directories added via --add-dir, /add-dir, or additionalDirectories are no longer automatically loaded into context at startup or after context compaction
• The tool rejects corrupted PNG images with a clear error message, preventing subsequent requests from failing when processing corrupted images
• Fixed processes hanging when the Bash tool failed to terminate properly

Hook Reliability Improvements and Safer File Encoding Handling

• Fixed SessionStart Hook context being lost after compaction
• The file editing tool now rejects writing unsupported text encodings, preventing silent rewrites that cause garbled text
• Added display of Hook batch progress and waiting time

Improved Session Reliability and More Slash Commands While the Agent Responds

• Fixed incorrect status display in the /plugins list when the same plugin is installed in multiple projects
• Fixed an issue where all tools from an MCP server were lost in the session when it failed to fetch the prompts or resources list
• Tightened proactive cross-session message delivery to reduce the chance of messages being mistakenly sent to other sessions
• Added support for executing more local slash commands directly while the Agent is responding, including /model, /fast, /effort, /skills, /agents, and plugin management commands

Performance optimization

• Improved session performance and runtime stability

Configurable session cleanup and faster /branch command

• Removed automatic session cleanup; it can now be enabled by modifying the configuration
• Improved the execution speed of the /branch command

Faster, More Reliable Session Management and Image-to-Image Generation

• Improved the execution speed of session forking and the /branch command
• File edits and writes no longer fail when a checkpoint backup can't be created, and rewinding now restores every file it can and clearly lists any files that couldn't be restored
• Added image-to-image support for the ImageGen tool

Use Sites with the qoder sites command

• Added support for using site capabilities via qoder sites subcommands

Windows sub-agent startup fixes and empty session exit cleanup

• Fixed being unable to start the sub Agent when output symlink creation failed on Windows
• Fixed unusable restore commands being shown when exiting an empty session

Improved Model List Loading and Update Experience

• Improved model list loading and update experience

Improved Model Response Reliability

• Fixed unfinished responses being incorrectly reported as completed when the model reached the output limit
• Improved automatic retry for Bailian model requests. When rate limits or concurrency limits are exceeded, requests now automatically wait and retry, reducing task interruptions

Improved MCP Connection Stability and Fixed Input Loss

• Improved MCP server connection stability
• Fixed typed and pasted content being lost while running /branch, /clear, and /new

Stability fixes for sessions, task cancellation, and Markdown rendering

• Fixed being unable to use ESC to cancel a generation task when the input box contains slash text; slash drafts are preserved after cancellation
• Fixed certain race conditions causing sessions to become unrecoverable
• Improved error messages and continuation handling when disk space is insufficient
• Fixed single tildes around numeric values and time ranges in Markdown rendering being misidentified as strikethrough and breaking bold text
• Improved handling of automatic compaction retry failures

Improved model information display

• Improved model information display

Built-in plugin marketplace and stability fixes

• Added the built-in qoder-plugins plugin marketplace
• Fixed an issue where using /btw after compaction affected the TUI status bar context usage calculation
• Fixed ACP session/load to return success only after all historical messages have been replayed.

Performance and web search improvements

• Improved web search results
• Improved startup speed and overall performance

Performance Improvements

• Optimized session performance and improved runtime stability.

Stability and reliability fixes

• Fixed a leak caused by failing to stop Windows background tasks
• Approval requests in Auto permission mode now follow the BYOK model

Improved Auto mode review flow and ACP tool status notifications

• Improved the two-phase review flow of Auto permission mode to speed up reviews
• Improved ACP tool status notifications

Goal Mode and Input Handling Improvements

• Improved Goal mode trigger rules so it is enabled only when explicitly requested
• Improved input compatibility in non-interactive mode

Chinese Input and Stability Fixes in Terminals

• Fixed an issue where Chinese characters were lost when typing in the integrated terminal of VSCode or VSCodium
• Fixed stability issues

Mermaid Diagrams in Markdown and /init Fixes

• Added support for Mermaid diagrams in Markdown rendering
• Press Ctrl+Enter during a model response to send queued messages immediately
• Fixed an issue where pressing Esc on the /init mode selection page did not cancel initialization
• Fixed an issue where the /init mode selection page always displayed AGENTS.md

Stability Improvements

• Fixed several issues and improved stability

Hardened Automatic Git Checks

• Fixed a security issue where internal Git calls could execute commands configured in a repository.

Auto Mode, Sites, and Bash Improvements

• Improved Auto mode: when the auto-approval component is temporarily unavailable, it now falls back to user confirmation and shows the reason
• Added support for setting site environment variables via the /sites management panel
• Added a warning for overly long AGENTS.md content
• Fixed an error that occurred during auto compaction
• Improved Bash command execution so configuration and plugin PATH are reliably passed through, with clearer bounded diagnostics for execution, snapshot, and output failures

Improved Configuration Handling and Skill Discovery

• Fixed concurrent configuration updates in Qoder CLI overwriting each other
• Improved skill discovery logic in the Skill tool
• Model requests now apply the model catalog’s default reasoning effort when no effort is configured, while preserving explicit settings and disabled reasoning.

Sites Publishing and Management with More Reliable Sessions

• Added Sites publishing and management, allowing websites to be created and published with natural language and viewed and managed in /sites
• Improved handling of compaction failures
• Improved Spinner to show Hook execution status event name and progress

Improved session name display and history browsing interactions

• Custom session names now appear on the input border, update immediately after renaming, and are restored when resuming or branching a session.
• Updated /btw history browsing with Shift+Arrow, [, ], and Tab shortcuts, contextual scroll hints, and protection against Shift+Tab accidentally changing the permission mode.

Improved the rename and btw commands

• Type /rename to auto-generate a session name based on the current conversation; manually entered names are automatically sanitized and length-limited
• /btw now keeps Q&A records within the current session and supports continuous follow-up questions; entering /btw again without a question reopens the latest answer, with options to browse, scroll, copy, or clear the records

Enhanced MCP connectivity and Chinese tool support

• Added support for connecting to MCP services via the Qoder MCP gateway
• Added support for recognizing and using Chinese MCP tools

Enhanced /model configuration and fixed preview and session resume issues

• Fixed the layout issue in the long content preview of the AskUser tool
• Personal edition users can now centrally configure and manage BYOK models with custom URL endpoints from the Custom page in /model
• Fixed being unable to resume history sessions when the Git repository is located at the Windows filesystem root
• Improved /model to support adjusting the model's thinking effort and context window, with all changes saved in a single confirmation

New video generation tool with performance and reliability improvements

• Fixed an issue in Windows Git Bash where a ShellSnapshot timeout could cause the first Bash command and the entire session to hang permanently
• Improved auto-compaction speed
• Improved the validation that requires reading a file before editing it
• Fixed date awareness becoming stale in long sessions spanning multiple days
• Added a video generation tool supporting text-to-video, first-frame-to-video, and reference-image-to-video; tasks run in the background and are automatically downloaded locally with a notification on completion

Clearer Permission Denial Feedback and Support Model Call Limits

• Improved the experience when a tool permission request is denied, adding clearer explanations that are easier to understand and follow
• Added support for setting a per-request limit on the number of model calls in TUI sessions

Faster Startup and More Flexible Model Selection

• Added the /output-style command to select an output style for the current project in the interactive interface
• Added support for applying a model only to the current session with s in the /model selector
• Upgraded the bundled runtime to Bun 1.4.2, improving startup speed, idle CPU usage, and steady-state RSS consumption
• Fixed special nested commands being incorrectly intercepted
• Fixed a file descriptor leak when a shell command fails to start after its output file has been opened.

Interactive Workflows and Chinese TUI Support

• Improved the workflow execution flow
• Added support for setting the TUI interface to Simplified Chinese through /settings
• Fixed path resolution errors when authorizing directories outside the workspace in Windows sessions
• Improved GLM context limit handling by automatically compacting history messages and retrying when conversations are too long

Performance Optimization

• Optimized performance issues

Stability Fixes

• Fixed stability issues

Remote Cloud Execution and Path Handling Fixes

• Fixed valid special paths being incorrectly blocked as UNC paths
• Fixed cloud task connection issues when created with '--remote'

Performance and Reliability Improvements

• Improved responsiveness and reduced CPU usage when resuming large sessions and rendering long streaming output
• Limited transcript loading in the session picker to improve performance with large session histories
• Fixed persisted scheduled tasks failing to run after session changes or when multiple sessions are open in the same project
• Labeled scheduled tasks owned by the current session in the OWNER column
• Increased the default /goal maximum interaction turn limit from 20 to 100
• Limited subagent spawning to one nested level by default
• Quest Skills can only be manually triggered by the user via the /quest command

Stability Improvements

• Bug fixes and stability improvements

Permission Mode Behavior Adjustments

• Adjusted permission validation behavior in Auto mode

External Providers and Reliability Improvements

• Supported correct Notification hook triggering during authorization prompts
• Fixed startup progress appearing in the session feed and canceled compaction leaving its progress indicator visible
• Fixed transcript lookup when Windows path casing differs
• Improved recovery after model output reaches the maximum token limit

MCP Authentication and Compaction Fixes

• Fixed MCP authentication expiration not being recognized correctly, and improved the pending-authentication hints in /mcp
• Fixed automatic compaction failing on oversized images

Permission System Fixes

• ACP sessions now offer the same selectable permission modes as the TUI
• Fixed several known permission issues

Permission System Consistency Overhaul

• The permission modes available over ACP now match those in the TUI
• Fixed several known permission issues

Terminal Setup and Input Experience Improvements

• Added the /terminal-setup command in the TUI: it automatically detects and configures the terminal's multiline input capability
• Improved inline command hints and completions
• Improved startup time

Stability Improvements

• Bug fixes and stability improvements

Skills Command Improvements and Context Accuracy

• Improved the skills command: -s and --scope are now consistently supported, and project can be used for the project-level scope
• Fixed skills install, skills link, skills uninstall, skills enable, and skills disable giving no feedback after running
• Improved the accuracy of usage statistics and categorization in /context
• Improved the usability of running /compact manually after switching models
• Fixed custom models (BYOK) being unable to handle images

UltraCode Mode and Startup Speedup

• Added UltraCode mode: Workflow multi-agent orchestration can be triggered with the /effort ultracode command or the ultracode keyword
• Fixed automatic compaction failing
• Improved the parallel tool execution experience: tools in the same parallel batch now show their status as soon as they finish

Auto-Compaction Accounting and Subagent Limits

• Fixed auto-compaction not being triggered
• Added a concurrency limit for subagents, configurable via the QODER_CLI_MAX_CONCURRENT_SUBAGENTS environment variable

Permission Coverage and Search Defaults

• Fixed directories configured with read deny rules still being searchable
• Fixed WebFetch deny/ask rules not taking effect
• Fixed settings.skills.disabled not applying to built-in bundled skills
• Fixed --max-turns not taking effect
• Fixed missing @-file completions and corrected case-sensitivity semantics

History Rewind After Compaction

• Fixed being unable to fork and rewind history messages after compaction

Skills Auto-loading for Added Directories and MCP Approval Fix

• Added automatic loading of skills in a directory added with /add-dir
• Fixed MCP tool calls being approved without a prompt in acceptEdits mode

Auto Mode Upgrade and Spell Checking

• Added an optional spellcheck setting that underlines misspelled English words as you type in the composer
• Improved the Auto mode classifier with a two-stage classification strategy, increasing classification speed and accuracy
• Improved the /mcp and /skills layout: entries are shown one per row, with descriptions displayed at the bottom, so more entries fit on screen
• Fixed the Auto mode classifier silently rejecting all subsequent authorizations after a certain number of consecutive rejections of dangerous actions
• Fixed the skills list being injected twice after resuming from a compaction
• Fixed uninstalled plugins being reinstalled by automatic plugin updates
• Fixed the failure message for an invalid PAT so it names the environment variable it came from and the priority that applied
• Fixed MCP tool result metadata and tool status display being lost

Goal Command, Tab Completion, and Worktree Improvements

• Added support for setting goals directly with /goal <objective>
• Improved slash-command Tab completion
• Improved session saving, restoration, and lifecycle management

Session Mentions, Settings Redesign, and Login Fixes

• Added @-mention of other live sessions in the composer, so you can reference a peer session alongside files and agents
• Improved the /settings layout: settings are shown one per row, with descriptions displayed at the bottom, so more settings fit on screen
• Changed /goal to no longer bind to Auto mode
• Fixed a stream truncated during the model's thinking phase not being retried, which interrupted the task
• Fixed login being blocked for a long time by a credential lock held by another process, and leftover locks are now recovered automatically with a retryable prompt

Tool Call and Session Resume Fixes

• Fixed compatibility issues with streaming tool call events on some models, improving the stability of parallel tool calls
• Fixed session resume taking too long in some worktree scenarios

Tool-Call Compatibility and Display Improvements

• Improved compatibility with some model service APIs, fixing some tool calls being skipped when tool_call_id values were reused
• Improved handling of host names containing non-ASCII characters, such as Chinese host names, with more accurate error messages
• Improved the TaskStop tool call card display, making the task stop status easier to read

Command Guardrails and Tool-Call Reliability

• Added parsing of wrapper commands such as xargs and watch in the Bash tool, so the real command is correctly identified and intercepted
• Fixed pkill -f possibly killing the CLI's own process in interactive sessions on Linux
• Fixed tool calls that take no arguments failing to execute in certain interruption scenarios under interleaved thinking, such as some MCP tools
• Fixed reasoning steps arriving out of order and losing their mapping in streaming output with interleaved thinking and parallel tool calls
• Fixed /goal set failing when the objective description contains command-line options
• Fixed scheduled upgrades still running after auto update was turned off

Plan Confirmations, Cross-Session Messages, and Long-Task Reliability

• Restored the confirmation dialog when entering and exiting Plan mode under auto and yolo modes
• Cross-session messages now show the sender and the message body inline
• Fixed shell task output files growing without limit, which could exhaust disk space
• Improved request timeout handling to resolve long-running tasks being interrupted incorrectly
• Improved diagnostics for abnormal responses, so real upstream errors are no longer hidden
• Improved the stability of multi-agent collaboration

Prompt Draft Stashing and Session Reliability Fixes

• Added Ctrl+S to stash the current prompt draft and restore it later
• Fixed teammate agents losing memory after being woken up
• Fixed EPERM errors when running /branch
• Fixed subagent tasks failing early when their state directory was not readable or writable
• Fixed prompt rendering glitches right after a session resume finished

Cross-Session Messaging and Scheduled Task Budgets

• Added cross-session messaging, so sessions can discover each other and send messages
• Added task budgets for scheduled tasks: set turn and credit limits with parameters, or change them in the panel
• Changed the default theme to auto, which picks the theme from your terminal background and keeps following it as the background changes at runtime
• Improved startup speed
• Fixed marketplace plugin installation failing
• Fixed input starting with an absolute path being misread as a command
• Fixed errors when switching to a non-vision model with image context
• Fixed automatic compaction using the wrong context window size

Cleaner Status Hints and Stability Fixes

• Improved memory saving notices to be less intrusive
• Added ESC to revert an accepted next-step suggestion
• Fixed the TUI possibly freezing when dragging in images that are slow to read or inaccessible
• Fixed some historical sessions failing to resume
• Fixed task failures caused by request interruptions during the model's thinking phase
• Fixed the AbortError crash caused by canceling a request

Status Line Enhancements

• Improved status line performance, with the option to display credits information
• Improved the startup guidance shown for the remote control daemon
• Fixed the subagent display in the TUI task panel
• Fixed /resume possibly hanging for up to 30 seconds

Auto Memory and Stability Improvements

• Added the auto memory switch, which can be enabled in the /settings panel or in settings.json. Once enabled, it automatically saves information worth reusing from conversations for later sessions
• Removed the length limit on custom system prompts passed with --system-prompt
• Improved the model queueing experience
• Improved login reliability

Agent Teams and Long-Session Reliability

• Added Agent Teams for multi-agent collaboration
• Improved task reminders in long sessions
• Fixed automatic context compaction not being triggered
• Fixed the Remote URL display issue
• Improved the Thinking status and Hook display interaction

Model Preferences and Progress Display

• Added per-model memory for reasoning effort and context window settings, which are restored automatically when you switch models
• Added remote ZIP plugin installation, and fixed the plugin status not updating after uninstall
• Added the security scan subcommand, which runs security scan tasks in Headless mode
• Improved the next-step suggestion feature, offering suggestions that better match the current conversation
• Improved session status hints, distinguishing between the Thinking and Generating reply phases
• Improved image paste responsiveness on Windows and macOS
• Fixed resuming a session whose worktree had already been deleted
• Fixed unreadable error messages when context compaction failed

Session Browsing and Scheduled Tasks

• Improved the /resume session browser: aligned shortcuts and clearer session metadata
• Improved the /crontab task panel: added a table header and clearer schedule interval descriptions
• Improved worktree stability
• Improved plugin loading logic
• Fixed duplicated image content when dragging multiple images at once
• Fixed session-level approval of plan mode not persisting

Display Improvements and Stability

• Improved the display pages
• Bug fixes and stability improvements

Service Account Auth and Session-Scoped Tasks

• Fixed persistent scheduled tasks being activated by unrelated sessions, and their owning session can now be managed in /crontab
• Goal creation is no longer globally unique: each session can have its own goal, and /goal take has been removed
• Entering and exiting plan mode no longer triggers a permission popup in Yolo and Auto modes
• The .agents/skills compatibility source is now enabled by default
• Removed the "Memorying..." progress indicator from the UI
• Improved the message shown when a streaming connection is interrupted
• Fixed remote artifacts failing to download on Bun binaries
• Added support for remote-controlled artifact display and download

Session Credits and OAuth Notice

• Added server-reported session credits usage display
• Added a manual authorization reminder before MCP servers start
• Improved keyboard shortcuts for external editor operations
• Improved Bash tool execution efficiency
• Fixed /clear and /new not resetting the persistent Bash working directory back to the startup root
• Added turn progress display in goal mode to improve the interaction experience

Rename, Skill Scanning, and Startup Improvements

• Added support for running /rename while a task is in progress
• Skill directories are now scanned recursively, with a visible warning when a declared directory contains no skills
• Improved startup speed by prewarming the shell snapshot and avoiding repeated loading of slow shell configs
• /logout now clears the login state immediately
• Fixed the external editor (Ctrl+X) failing to open in some integrated terminals such as Cursor
• Fixed input conflicts while a GUI external editor was open — the terminal is no longer handed over
• Improved the Ctrl+O detailed view toggle logic
• Improved /compact command error messages

Loop Scheduling Upgrade and Workflow Improvements

• Upgraded /loop: the agent can now decide wakeup intervals on its own, with support for task persistence
• Added the /crontab command dashboard for viewing and managing scheduled tasks
• Added inline ghost-text completion for skill and slash commands in the input box
• The TaskStop tool now supports stopping running workflows
• Improved the Workflow TUI panel display logic
• Fixed false stall warnings while workflow agents were streaming
• The default permission mode in /settings can now be selected via options
• Fixed known issues with slash command parsing and completion
• Fixed the terminal title being lost on Windows
• Fixed skill directories failing to load when a root-level SKILL.md was present

Prompt Suggestion Toggle and Stability Fixes

• Added a settings toggle to enable or disable prompt suggestions
• Added the .agents/skills compatibility toggle to /settings
• Fixed slash commands being reported as unknown before the command catalog finished loading, and reduced redundant catalog rebuilds
• Showed the full Bash command in the expanded shell output view
• Improved Markdown table row boundary rendering
• Fixed UI freezes when resuming long sessions
• Fixed Monitor notifications not being delivered in time
• Fixed Grep/Glob tool cross-platform compatibility issues

Subtask Command, Feedback Enhancements, and Configurable Directories

• Added the /subtask command for dispatching background subtasks from the current session: the subagent inherits the session context, runs asynchronously, and reports back its result.
• Added /feedback argument support to prefill feedback content
• Added settings configuration to control skill loading from the .agents directory
• Adjusted the default maximum turns of local subagents to 150
• Fixed mobile conversation display issues
• Fixed long MCP user config paths not being fully displayed

MCP Enhancements, Task Management, and Reliability Fixes

• Improved plugin validation and update logic
• Improved Ctrl+B shortcut compatibility
• Added persistent task management
• Improved WebFetch timeout error surfacing
• Adjusted model display name in headless mode
• Improved command-hook process tree teardown

Background Tasks, Terminal Titles, and Input Enhancements

• Improved background task control with monitoring, stopping, and clearer status feedback
• Added the useProxyFromEnvironment setting to toggle reading proxy configuration from environment variables via /settings
• Improved terminal session window title updates
• Added Ctrl+K / Ctrl+U / Ctrl+W line-editing shortcuts in dialog and settings text inputs
• Improved the /resume session picker
• Surfaced argument parsing errors on stderr and aligned unknown-option output
• Fixed the subagent thinking depth setting
• Fixed image reading failures on Linux

Compression, MCP, and TUI Improvements

• Added compression progress display in the TUI to improve the interaction experience
• Added configuration for shell command model responses, disabled by default
• Added timeout control for MCP tools
• Improved permission behavior for referenced files
• Fixed MCP list names being collapsed
• Fixed the viewport not restoring after the suggestion list closes

Workflow Refinements

• Improved workflow pause and resume logic
• Improved Deep Research delivery and queued task dispatch
• Improved startup header and update notification display
• Improved prompts on model refusal

Improvements

• Improved error message display
• Enhanced code security scanning reliability

Security Scan & Model Management

• Officially released native code security capabilities
• Support refreshing the model list on demand via Ctrl + R

Shell & TUI Experience, and Fixes

• Optimized !bash mode execution and output behavior
• Improved the built-in Deep Research workflow
• Optimized underlying network requests to improve reliability and stability

Command Execution, Shell Experience, Exit Cleanup, and Hook Fixes

• Optimized UI rendering during command execution
• Optimized the auto-allow detection logic for read-only shell commands
• Cleaned up unfinished background tasks when the CLI exits
• Fixed an issue where the PreToolUse hook did not take effect in headless and SDK modes

Input Experience and Stability Improvements

• Optimized CLI terminal rendering logic
• Ensured subagents honor the disallowed tools configuration
• Optimized memory usage

Interaction Experience and Startup Stability Improvements

• Improved TUI rendering stability
• Improved the interaction experience for subagent, feedback, remote control, and other features
• Fixed some known startup blocking issues

BYOK configuration and rendering improvements

• BYOK models now support context window and reasoning depth adjustments.
• Added a Plan mode toggle to the /settings dialog.
• Improved streaming rendering performance by caching stable Markdown blocks.
• Fixed color display anomalies between TUI rendering frames.
• Reduced the impact of the status bar marquee on usability.
• Improved worktree stability.

Thinking stability enhancements

• Improved the stability of model thinking processes.

Startup performance and worktree experience improvements

• Optimized the startup mechanism for faster launch.
• Improved the worktree experience.

Workflow, Skills, and permission improvements

• Improved the workflow orchestration experience.
• Added support for installing remote Skill packages over HTTPS.
• Refined the detection scope for redirect commands in the Bash tool.
• Improved error messages.
• Improved the tool-calling experience.

Custom models and model selector enhancements

• BYOK models now support custom model names.
• Improved product interactions and onboarding experience.

Feedback image attachments and experience improvements

• Added image attachment support to the /feedback dialog.
• Improved the exit options for Plan mode.
• Streamlined permission checks for the worktree tool.
• Added a toggle in /settings to enable or disable code usage statistics.

Optimized status bar and model prompts

• Improved UI notification display, including status bar prompts and model list prompts.

Optimized input and display

• On macOS, you can now delete pasted images as a whole block.
• Reduced the impact of status bar promotional messages on scroll control.
• Fixed the display of context usage when resuming a compressed session.
• Improved the default effort label display in the model list.

Improved stability and user experience

• Fixed an issue where a tool's oversized single-line output degraded TUI rendering performance.
• Improved plugin marketplace management.
• Improved TUI display and refined some error messages.

Plugin marketplace and permission mode improvements

• Added support for the plugin marketplace.
• Upgraded the Qoder CN CLI rules type to support both autonomous model decisions and manual inclusion.
• Fixed an issue where the Plan and Ask tools were not correctly disabled in headless mode.
• Fixed an issue where AskUser was silently rejected in auto mode.
• Fixed an issue where the ECS key could not switch from INSERT to NORMAL during streaming output in Vim mode.
• Fixed an issue where secure slash commands did not respond when idle.

Enhanced default permission mode configuration and improved user experience

• The default permission mode now supports `yolo` and `dontask` configurations.
• Fixed an issue where interruption records were rendered repeatedly in the TUI.
• Improved the display of the Background Agent.

Improved Plan mode and model experience

• Added the `--config-dir` parameter to support custom configuration directories.
• Fixed an issue where Plan mode did not support Shell read commands.
• Improved the stability of model requests.

Improved stability and UI experience

• Improved API error messages.
• Improved exception handling for tool calling.
• Improved the UI display and user experience.

Enhanced dynamic workflows and memory function

• Added support for the dynamic workflow feature.
• Added support for `.qoder/rules/**.md`.
• Added user-level Memory and improved the Memory experience in Plan mode.
• Improved the startup speed of Headless mode.

Improved network proxy

• Improved the network proxy mechanism. It now supports HTTP_PROXY, HTTPS_PROXY, and SOCKS5 proxies.

Improved startup performance and model reliability

• Added configuration items for first-packet timeout and stream idle timeout for model Server-Sent Events (SSE) responses.
• Improved the context window backoff policy for models.
• Improved the timeout for the SessionEnd Hook and the failure shutdown policy for PreToolUse.
• Hook validation: Added the required `hookEventName` field to `hookSpecificOutput`.
• Improved the performance of the SDK when retrieving the model list.

Improved model list caching

• Improved the model list caching mechanism.

Improved permissions and command stability

• Fixed an issue where commands were truncated.
• Fixed an issue where Git-related environment variables were cleared during command execution.
• The `/resume` command now adheres to the current permission mode.

Improved welcome page and fixed stability issues

• Improved the welcome page display by adding a marketing and promotion module and model discount notifications.
• Fixed a parsing failure caused by redundant parameters in the Read tool.
• Refactored the download and installation process for Windows Git Bash to improve stability.

Workflow orchestration

• Improved the notification experience for background tasks.
• Fixed an issue with the speech input termination process.

Improved model panel experience and fixed session recovery

• Improved the model selection page.
• Fixed an issue where interruption records were not displayed after using `resume`.

Improved session interruption and permission experience

• Added a session interruption feature. Messages entered during a conversation are now sent immediately.
• Enriched the authorization information displayed in the `/permissions` command panel.
• Improved the loading performance when using `/resume` for very large sessions.
• Fixed an issue where the permission pop-up was incorrectly auto-approved after switching modes.

UI improvements

• Improved the display style of Hook messages in the session stream.
• Improved the interactive experience for model parameter settings.
• Improved the tool permission reminder logic for certain paths.
• The input box content is now retained when you press ECS to exit a terminal dialog.
• Removed the timeout limit for permission requests on the ACP side.

Enhanced Skill commands and models

• Added the built-in skill commands `/run`, `/verify`, and `/run-skill-generator`.
• Added the `/effort` command.
• Aligned the Subagent frontmatter configuration with the existing permission system.
• Fixed an issue where the image recognition tool failed in Node SEA products.

Fixed context warning threshold

• Fixed inconsistent context warning thresholds.

Queue progress and issue fixes

• Added a model queue progress bar chart that shows the elapsed wait time and the estimated maximum wait time.
• Improved the auto compression experience.
• Fixed an issue where the `/context` and status bar percentages were inconsistent in long sessions.
• Improved the display style of Subagents.
• Fixed an issue with the display of error messages when an MCP tool calling exception occurred.
• Improved error detection and prompts for excessive input length.

Improved TLS trust, TUI, and stability

• Added trust for the operating system certificate store and support for custom CA certificates.
• Improved the display of Subagents using event-driven transcript rendering.
• Fixed TUI support for `sendString` input in IDE integrations.
• Improved compatibility with the `screen` terminal multiplexer.
• Performance optimization for `/plan`
• Improved the TUI display performance when resuming large sessions.
• Fixed the Bash stdin redirection behavior.
• Added an error message for HTTPS proxy TLS handshake failures.

Agent sessions enabled by default, Permissions upgrade, and SOCKS5 proxy

• The Agent session feature is now enabled by default.
• Upgraded the `/permissions` command with directory completion, a workspace dialog box, scope-colored labels, and a delete confirmation.
• Added support for SOCKS5 proxy.
• The AskUser dialog box now supports CJK input, and an issue with the cursor reset has been fixed.
• Refactored the background task details panel for structured display and fixed an issue with empty output.
• Plugin hooks are no longer affected by the `--setting-sources` filter.
• Enhanced `asyncRewake` hooks with support for custom messages, automatic fallback in non-interactive mode, and prefix matching.
• `/memory` no longer exposes the actual file path pointed to by a symbolic link.
• Ownership is now restored for `/goal` across `resume`. The goal and permission mode are reset when using `/clear` or `/new`.
• The restored session history is now retained when switching models or inference strengths.
• Fixed a mismatch between MCP environment variable VAR expansion and OAuth discovery resources.

Stability improvements

• Fixed a dependency issue in the npm version.

Plugin path placeholders and issue fixes

• Added `PLUGIN_ROOT` and `PLUGIN_DATA` path placeholder substitution in plugin content.
• Fixed an issue with duplicate warnings when switching permission modes.
• Fixed an issue where project permission rules did not take effect immediately after granting workspace trust.
• Fixed an issue with incorrect control character parsing on Windows.
• Improved startup speed.

Standalone Plan mode and stability improvements

• Plan mode is now independent of the permission loop. Use `/plan` to enter and exit. Shift+Tab now only cycles through permission modes.
• Added a "Start goal execution" option when exiting Plan mode to seamlessly transition to the Goal flow.
• Fixed an issue where a session would permanently freeze if a Hook child process timed out or was aborted.
• Fixed an issue where Goal and Plan states were not correctly restored after a process restart.
• Fixed an issue with incorrect modules in the npm release package.
• Improved the prompt text for switching permission modes and the options for exiting Plan mode.

Model list, activity recommendations, and plugin improvements

• Added the `--model-list` parameter to enumerate available models without a UI.
• Added polling for the model service queue status. The service automatically resumes when it becomes available again.
• Fixed an issue where `Agent(name)` deny rules in the permission configuration were not effective.
• Fixed an issue that occasionally caused session recovery to fail.

Improved YOLO mode and startup performance

• In YOLO mode, `EnterPlanMode`/`ExitPlanMode` no longer triggers a pop-up, and `AskUserQuestion` can be passed through normally.
• Added support for customizing automatic mode classification rules in the settings.
• Fixed a voice reconnection issue to avoid expired WebSocket close events and repeated activation loops, improving stability.
• Improved the experience of launching a browser to log on in Windows.
• Isolated Goal by session to fix crosstalk issues between multiple windows.
• Improved TUI display stability during prompt input completion.
• Reduced the performance impact of automatic updates.

TodoWrite and UI improvements

• Improved the TodoWrite validation recovery mechanism and error messages.
• Unified the color scheme for info messages and fixed the display color of inline code.
• Cancellation prompt messages are now displayed in gray to improve visual hierarchy.
• Fixed several other issues.

Voice polishing and stability improvements

• Added a voice transcription polishing feature to improve input accuracy.
• Improved the experience for Bring-Your-Own-Key (BYOK) providers. Disabled states now show explanations, and model identifiers are clearer.
• Improved the submission confirmation interaction on the Review page.
• Improved the permission denied prompt in YOLO mode.
• Fixed several known issues.

Revamped YOLO mode and permission improvements

• YOLO mode now automatically approves all tool calls without requiring pop-up confirmation.
• The `--tools` parameter now supports multiple values.
• Improved the hierarchy and clarity of the feed stream display.
• Fixed an issue on Windows where a console window would pop up during child process execution.
• Fixed an issue where pasted content in the prompt input was incorrectly parsed as an `@` reference.
• Improved the loading display for Skills and Hooks.

Loop and Hooks are now generally available

• The `/loop` command and scheduled task tool are now generally available.
• The `/hooks` command is now generally available.
• Added support for configuring startup parameters through environment variables.
• Custom Agent names are no longer restricted to the slug format.
• Removed the quantity limit for the `@agents` completion list.
• Improved the status indicator, message spacing, and visual presentation of the chat interface.
• Fixed several other issues.

Performance and stability improvements

• Improved startup performance by reducing startup time through delayed initialization and resource caching.
• Added a TUI interface snapshot to the feedback diagnostic package to provide more complete context information.
• Fixed multiple issues and optimized interactions.

Image search, streaming speech input, and enhanced plugin capabilities

• Added the built-in ImageSearch tool to search for images directly within the Agent.
• Added streaming speech-to-text transcription. Recognized text now flows directly into the input box.
• Improved the experience of the plugin browser for items pending removal and in empty Marketplace scenarios.
• Fixed several other issues and made optimizations.

Stability and experience optimizations

• Added a banner to indicate when the quota is exhausted.
• Improved the stability of network requests and model data transmission.
• Fixed an issue where speech input would terminate recording immediately after disconnection.
• Fixed a parameter schema parsing failure for built-in tools with Qwen series models.
• Fixed several other issues.

Stability improvements

• Fixed a runtime exception in the npm package caused by a tagged template encoding error.

Voice, Remote, and interface fixes

• Fixed an issue where transcribed text flickered during voice recording and changed the behavior to insert the transcribed text at the cursor position.
• Improved the performance of Agent task execution.

Goal, voice, and stability fixes

• Fixed stability issues related to speech input (recording tool detection, termination timeout, multi-segment input handling).
• Improved the stability of the model calling service.
• Fixed an issue where the working directory was not correctly synchronized in Worktree sessions.
• Fixed an issue where a permission confirmation would incorrectly pop up when accessing project configuration files in automatic permission mode.

Voice and stability fixes

• Added the ElicitationResponse Hook notification type.
• Fixed stability issues in voice mode.
• Fixed an issue where session title generation failed when the compact model was unavailable. It now automatically falls back to the chat model.
• Fixed a race condition at startup in the BYOK model directory.

Introducing Qoder CN CLI 1.0

• Cloud runtime: Added support for the `qodercli --remote` flag. Cloud mode lets you submit tasks to a cloud container for one-time execution. The workflow continues to run for hours even if you shut down your machine or lose your network connection.
• RepoWiki: Added support for the `qodercli wiki` subcommand to generate repowiki documents locally without your code leaving your domain.
• Voice: Added support for the `/voice` command. The speech input feature is fully upgraded and available for free.
• Goal: Added support for the `/goal` command to keep long-running, goal-oriented tasks on track.
• Open model parameters: You can now customize `efforts` levels and set a 1M context window.

• Agent SDK: Supports TypeScript and Python. Integrate full Agent capabilities with just a few lines of code. Provides programmatic control, including tool authorization callbacks, pre-execution interception, and parameter modification.
• Cloud Agents: Agents as a Service. Built-in features include sandbox isolation, state management, asynchronous orchestration, credential management, and a long-running platform.

• Orchestration: Tasks are automatically split and executed in parallel by multiple Agents. The main Agent determines the splitting policy, and Subagents run in independent event loops.
• Extensions: Supports five orthogonal extension points: Skills, Hooks, MCP, Subagent, and Command.
• Autonomy: The Agent assesses risks on its own and only requests confirmation for critical nodes. Automatically routes requests to Large Language Models (LLMs), Small Language Models (SLMs), or Vision Language Models (VLMs).
• Integration: Supports three integration modes: Headless (for unattended CI/CD), ACP (for standard bidirectional communication), and SDK (for TypeScript/Python).