Invite a friend, get 1 month of Pro freeHow it works
AgentVera
EN
Download

Qwen Code changelog

The last 18 stable releases of Qwen Code, newest first, with the lines that remove, rename or deprecate something pulled out on top. Taken from the GitHub releases every morning.

Latest: 0.26.4 · October 29, 2026 GitHub → About Qwen Code →

Breaking changes and deprecations

Latest releases

  1. 0.26.4

    • feat(generated-ui): follow the host theme in generated views
    • feat(ci): bound autofix takeover with a round budget
    • feat(verify-pr): make "re-verify after update" a follow-up trigger
    • feat(extensions): support unscoped npm installation and updates
    • feat(web-shell): move no-workspace sessions beside workspaces
    • feat(managed): add independent worktrees for main sessions
    • feat(web-shell): allow embedding hosts to pick workspace directories
    • feat(generated-ui): localize and reshape the card frame
    • feat(web-shell): show model timing inline in Token Usage
    • feat(cli): raise workspace skill install limits
    37 more on GitHub →

  2. 0.26.3

    • Generated UI cards keep inputs and state across navigation and split views, show live views after initial calculation, and appear only for clients that can render MCP Apps.
    • Persistent session Workflow mode can be enabled with --workflow-mode on, /workflows mode on, SDK, or ACP, and workflow switches now reach the CLI sandbox.
    • Tool results shrink to remaining headroom before auto-compaction, and qwen-latest-series- or qwen4 aliases are detected with 1M-token context windows.
    • Valid XML tool calls are recovered when parameters quote other call markup, malformed blocks are isolated, and quoted examples are not dispatched.
    • ACP and daemon clients keep rewind identity consistent across retries, cancellations, checkpoints, recordings, and notifications, while SDK frames preserve subagent identity.
    • /autofix-pr start stops when the PR head branch has unpushed commits or uncommitted changes, preventing the watch from overwriting unseen work.
    • feat(memory): opt-in extraction cadence after a no-op run
    • Shrinks tool results to remaining headroom before auto-compaction, respecting explicit truncation settings and protected content to reduce unnecessary compaction.
    • Keeps eligible Generated UI cards alive across history navigation, session switching, and split-view changes in the same page, preserving inputs, controls, and runtime state.
    • Adds persistent session Workflow mode through --workflow-mode on, /workflows mode on, SDK, and ACP controls, with saved mode shown in Ink and OpenTUI.
    32 more on GitHub →

  3. 0.26.2

    • fix(managed-agent)!: simplify Web Shell controls and retire cwd submissions
    • feat(managed): add session permission rules and Auto approval
    • feat(core): add an experimental shared agent loop and run headless on it
    • Adds X-DashScope-Source-Config to matching QwenCloud and Qianwen AI Platform requests to identify site and client, without changing authentication or retries.
    • Resumed OpenTUI sessions now retain recorded text, images, tool results, and prompt identities, while rewind restores transcript, request, and draft.
    • OpenTUI now captures complete checkpoints at edit approvals and restores them safely, preserving conversation metadata and blocking stale or cancelled owners.
    • Adds automation.scanner-threads to run managed automation scanner rows on a dedicated pool and leases rows from claim time.
    • Adds Continue analysis with current parameters so Generated UI users can review and send frozen parameters to the original session.
    • Adds an opt-in external-scratch Runtime provider that lets an operator-run HTTP service place Session workers while preserving attestation and recovery.
    • Adds a prompt argument hint for /verify-pr so users can see the PR number and supported flags such as --publish and --watch.
    146 more on GitHub →

  4. 0.26.1

    • fix(qqbot)!: restore per-group session isolation under thread scope
    • Adds opt-in daemon token persistence: qwen daemon start --save-token stores the bearer token in an owner-only file, qwen daemon token manages it, and daemon stop removes it.
    • Adds idle-only manual context compression for Workspace-bound Managed Sessions via public API and the WebShell context-usage popover, preserving audit history.
    • Aligns OpenTUI input, editing, drafts, and /statusline behavior with Ink, adding configurable status lines and more predictable completion, queue, and cancel handling.
    • feat(shell): remind the model to record files a command writes
    • Managed Workspace recovery now waits for durable settlement before continuing, shows recovery reasons and valid actions in WebShell, and gates message sending.
    • Embedding hosts can now show unavailable models as disabled rows in the composer picker, with hover or focus hints and optional host-provided actions.
    • feat(managed-agent): Add session lifecycle UI and reading exports
    • feat(generated-ui): expose interaction outcomes and applied parameter snapshots
    • feat(managed-agent): expose durable capacity observations
    146 more on GitHub →

  5. 0.26.0

    • feat(daemon): support session-targeted Skill and Extension refresh
    • feat(runtime-broker): Session-create warm for Workspace-isolated bound Sessions
    • feat(managed-agent): ship a configuration template in the reference distribution
    • feat(serve): run private CSI file turns through Hosted Harness
    • feat(managed-agent): support local and SSH workspace file references
    • Adds a read-only --check preflight that prints PASS, WARN, FAIL, or SKIP for managed-agent deployment checks and exits nonzero on failure.
    • Lets a successor Hosted owner finish private CSI file turns that already have durable receipts, preserving tool responses and avoiding repeated file operations.
    • Managed Agent submissions now accept up to 8 MiB of attachment input, preserve large pending requests across refresh, and let users discard unrecoverable drafts.
    • Hosted Turn recovery now asks the Runtime Broker for execution evidence and treats unprovable outcomes as unknown instead of transient faults.
    • Adds systemd units and setup guidance for single-host managed-agent deployments so Runtime workers stay running when Broker restarts
    227 more on GitHub →

  6. 0.25.3

    • Adds optional model-generated layouts to Test Failure Explorer, with validation and fallback to the fixed view
    • The Web Shell mobile composer no longer shows the ↑ and ↓ history buttons, reducing clutter while keeping history search available.
    • The Web Shell trajectory panel now includes a collapsible Recent executions section that shows live requests and tools with elapsed time and final durations.
    • Managed sessions created from published AgentDefinitions can now load static Skills and pass pinned MCP server and Hook catalog selections to Hosted Runtimes.
    • Files now offers New file and New folder actions, and folder Delete asks for a second confirmation before permanently removing the folder and its contents.
    • Eligible legacy cancellations now require confirmation through the daemon or WebShell Continue flow before continuing, while explicitly user-cancelled work stays blocked.
    • Hosted parent agents can launch snapshot-isolated child agents that read a frozen workspace copy without changing files, running commands, or affecting the parent.
    • Clicking the workspace folder beside the chat title now opens and selects the Files tree in the right panel, with keyboard and focus preserved.
    • Managed sessions now integrate into the home and sidebar, supporting workspace creation, approval policy and model selection, and context usage in the composer.
    • The Files panel adds a Delete menu for regular files, requiring a permanent-deletion confirmation for the selected trusted local workspace.
    179 more on GitHub →

  7. 0.25.2

    • feat(managed-agent): H4e-b2 the lead's mailbox over Session messages
    • A new private extension imports Vitest 3.2 JSON reports and exposes failures, errors, counts, and diagnostics through MCP tools with search and pagination.
    • Opt-in headless Mods now support TypeScript authoring, relative imports, typed non-secret configuration, and dependency validation before startup.
    • Adds a default-off option allowing authorized public REST and WebShell sessions to use foreground Shell with mandatory approval.
    • feat(managed-agent): Introduce the EventTransport seam behind the after-commit publisher
    • feat(managed-agent): Stop Hosted team members with request_shutdown (H4e-b3a)
    • Opt-in headless Mod runtime gains ordered middleware dispatch, command interception, and bounded failure recovery without resetting module state.
    • Adds a Test Failure Explorer app that lets users browse, search, page through, and copy analysis requests for up to 20 selected test failures.
    • Enables H3 Monitor and background Shell capabilities, allowing background runs to outlive their turn and output artifacts.
    • Adds a Java mirror of the managed-event-envelope contract with derivation, validation, deduplication, and fixture tests, without changing runtime behavior.
    163 more on GitHub →

  8. 0.25.1

    • Restores the selected text-only connection and validated session or workspace route after Android Activity recreation, while preserving explicit retry screens after failures.
    • Improves mobile profile screen accessibility with clearer connection actions, editor labels, validation announcements, and scrollable storage recovery.
    • Adds an experimental Kubernetes CSI-backed runtime that validates workspaces, persists pod identity, and records durable worker acknowledgements for managed sessions.
    • Adds background Shell and Monitor capabilities to the Managed Agent runtime, enabling durable background shell sessions and task event reporting.
    • Web Shell /btw side can now create independent side-task conversations in a parent session's secondary workspace, including while the parent is responding.
    • Agent and Goal coordination tools are now deferred by default and discovered on demand, reducing unnecessary tool declarations in ordinary tool-call sessions.
    • Local Managed Agent runtime tool outcomes are now durably recorded, including parameters, receipts, and checkpoints, before results are returned to the model.
    • Hosted approval requests can now include bounded previews of file and shell input, exposed through public and WebShell APIs with truncation metadata.
    • Lets creators change a Workspace-bound Managed Session's working directory through a durable, idempotent operation.
    • Adds read-only glob file discovery to Hosted Workspace /2 profiles while keeping searches inside the session working directory.
    179 more on GitHub →

  9. 0.21.2

    • Autofix now defers lower-severity suggestions after five rounds and posts visible notices when refusing to proceed due to round limits.
    • Added an opt-in Auto Recall profile that automatically searches external context with built-in security limits.
    • Enhanced the /verify command with increased time budgets, evidence screenshots, and new maintainer verification techniques.
    • Web Shell now supports contextual task panels, theme-aware visuals, and correctly handles artifact previews and session states.
    • GitHub channels now emit exactly one final comment per event and display eyes reactions to indicate active agent processing.
    • Added APIs and UI controls to view, approve, and revoke channel pairing approvals within specific workspaces.
    • Autofix now defers lower-severity suggestions after five change rounds, allowing only critical findings and failed checks to drive further code modifications.
    • Goal v3 is now connected to the turn engine with deterministic permit handling, ensuring runtime continuations do not consume session or recursive turn budgets.
    • Static image reads now return a canonical, auto-oriented JPEG overview with source dimensions to support reliable zooming within shared resource budgets.
    • The triage system now mechanically triggers the 2b-bis sandboxed-lane recommendation when a Stage 2 draft explicitly contains a not-verified sentence.
    95 more on GitHub →

  10. 0.16.1

    • chore(release): v0.16.0
    • fix(core,cli): close tool_use↔tool_result invariant across all failure paths
    • fix(vscode): skip redundant tsc build in prepackage to prevent TS5055
    • fix(core): preserve tab-indented notebook formatting
    • feat(telemetry): Phase 4a — TTFT capture + GenAI semconv dual-emit
    • fix(scripts): renormalize CRLF storage for install-qwen-standalone.bat
    • fix(build): tree-shake React reconciler dev build to prevent PerformanceMeasure leak
    • chore(deps): update express from 4.21.2 to 5.2.1
    • fix(cli): stabilize flaky sticky-todo remeasure test
    • fix(cli): gate mintty OSC 8 detection on TERM_PROGRAM_VERSION ≥ 3.3
    1 more on GitHub →

  11. 0.16.0

    • Breaking ci(deps): bump docker/ actions to Node 24 majors (silences GitHub Node 20 deprecation warning)
    • feat(cli): wrap markdown links in OSC 8 so wrapped URLs stay clickable
    • fix(core): normalize cumulative OpenAI stream deltas to suffixes
    • fix(cli): auto-restore prompt and preserve queue on cancel
    • docs(auth): add custom API key wizard PRD
    • feat(cli): support batch deletion of sessions in /delete
    • fix(core): tag subagent OpenAI JSON logs
    • feat(subagents): use fastModel for Explore subagent
    • fix(dashscope): use URL hostname check instead of regex to avoid ReDoS (CodeQL)
    • fix(core): improve runtime fetch options error handling and documentation
    • fix(telemetry): address PR #3847 review follow-ups for trace correlation
    97 more on GitHub →

  12. 0.15.11

    • chore(release): v0.15.10
    • perf(core): bound session-list metadata reads to head/tail 64KB; pool buffer; lazy message count
    • test: stabilize main e2e flakes
    • feat(cli): core built-in i18n coverage
    • feat(core): write runtime.json sidecar for active sessions
    • feat(telemetry): inject traceId/spanId into debug log files for OTel correlation
    • feat(tools): defer low-frequency built-in tools to reduce initial prompt size
    • ci: skip unnecessary release and SDK checks
    • feat(installer): add standalone archive installation
    • feat(cli): Ctrl+B promote keybind (#3831 PR-3 of 3)
    24 more on GitHub →

  13. 0.15.10

    • chore(release): v0.15.9
    • fix(cli): validate /model command arguments
    • fix(core): log the OpenAI request actually sent on the wire
    • fix(core): drop disabled MCP server from health status registry
    • test(cli): drop wait-dependent SessionPicker search tests (closes
    • feat(core): add reactive compression on context overflow
    • fix(core): filter Mistral reasoning content at request boundary
    • fix(cli): preserve comments and formatting in settings.json during migration write-back
    • [codex] fix monitor notifications for subagents
    • feat(memory): add autoSkill background project skill extraction
    11 more on GitHub →

  14. 0.15.9

    • feat(telemetry): add sensitive span attribute opt-in
    • chore(release): v0.15.8 [skip ci]
    • feat: add commit attribution with per-file AI contribution tracking
    • fix(core): per-agent ContentGenerator view via AsyncLocalStorage
    • refactor(cli): provider-first auth registry with unified install pipeline
    • fix(core): accept partial reads in prior-read enforcement
    • fix(cli,core): live-phase panel-ownership filter + post-delete statusChange emit
    • fix(core): close bound-tool gap on runForkedAgent's YOLO wrapper
    • feat(sdk-python): replace verbatim release notes inheritance with --generate-notes
    • ci(release): keep skip-ci out of release PR titles
    12 more on GitHub →

  15. 0.15.8

    • chore(release): v0.15.7
    • test(sdk): align tool-control E2E with prior-read enforcement
    • fix(skills): allow symlinks pointing outside the skills directory
    • feat(web-templates): add light theme and toggle to /export HTML
    • feat(cli): replace inline AgentExecutionDisplay with always-on LiveAgentPanel
    • ci(issue-followup-bot): render bot comment newlines correctly
    • ci(release): skip CI on the version-bump squash commit on main
    • fix(core): foreground agent entry lingering in status bar after completion
    • fix(cli): prevent ESC in background tasks dialog from cancelling running request
    • fix(memory): address code review feedback for auto-memory recall
    1 more on GitHub →

  16. 0.15.7

    • chore(release): v0.15.6
    • feat(core): add FileReadCache and short-circuit unchanged Reads
    • fix(cli): honor proxy setting
    • chore: remove legacy Gemini workflows
    • feat(core): add shared permission flow for tool execution unification
    • fix(cli): restore SubAgent shortcut focus
    • Add background agent resume and continuation
    • feat(review): expand review pipeline + qwen review CLI subcommands
    • feat(telemetry): define HTTP OTLP endpoint behavior and signal routing
    • fix(vscode-companion): align package eslint config with root and style cleanup
    47 more on GitHub →

  17. 0.15.6

    • fix(memory): use project transcript path for dream
    • fix(cli): bound SubAgent display by visual height to prevent flicker
    • fix(cli): keep sticky todo panel compact
    • fix(core): replay DeepSeek reasoning_content on all assistant turns
    • fix(cli): correct model precedence — argv > settings > auth env vars
    • fix(core): preserve reasoning_content in rewind, compression, and merge paths
    • fix(cli): persist directory add entries
    • fix(lsp): 修复 LSP 文档、isPathSafe 限制，并提升 LSP 工具调用率
    • fix(vscode-companion): fill slash commands into input on Enter instead of auto-submitting
    • fix(ci): add merge-back PR for stable releases in release workflow
    1 more on GitHub →

  18. 0.15.5

    • mcp config as cli
    • fix(cli): refresh static header on model switch
    • feat(core): wire background shells into the task_stop tool
    • feat(skills): add tmux-real-user-testing skill for readable TUI test logs
    • feat(cli): wire background shells into combined Background tasks dialog
    • fix(core): inject reasoning_content on DeepSeek tool-call replays
    Release notes →

Other agents

Claude CodeCodex CLIGemini CLICopilot CLIopencodeHermes AgentKimi CLIGooseMistral VibeCrushAiderCline All agents →

Run Qwen Code next to the other agents and see when one needs you: AgentVera keeps them in one window on Mac, Windows and Linux, and on your phone.

Bring your agents to one desk.

Download, pick a project, open your first agent. Your installed CLIs are ready to go.

Download for free