Tools / Qwen Code / v0.28.0

Qwen Code v0.28.0 release notes & breaking changes

breakingsecuritymodel

Qwen Code v0.28.0 is a Alibaba release published 2026-11-17 with 110 listed changes, flagged for breaking changes, security fixes. Summary and full notes below, extracted from the official notes.

Get breaking changes for Qwen Code in your inbox every Monday

One free email: every breaking change, retirement and price change across 23 coding agents. No spam, unsubscribe anytime.

Subscribe free

← All Qwen Code releases and breaking changes

Summary

Full notes

<!-- qwen-release-notes:v1 -->

## Highlights

_See the complete change list below._

## Breaking Changes

No known breaking changes.

## Complete Change List

### Features

- feat(web-shell): add sidebar workspace filter ([#17244](https://github.com/QwenLM/qwen-code/pull/17244)) by @callmeYe
- feat(managed-agent): add the DingTalk adapter on the managed channel path ([#17136](https://github.com/QwenLM/qwen-code/pull/17136)) by @wenshao
- feat(acp): route live and read-only Agent Host sessions through the shared tool batch ([#17218](https://github.com/QwenLM/qwen-code/pull/17218)) by @yiliang114
- feat(core): let the loop switches be forced off, not only on ([#17048](https://github.com/QwenLM/qwen-code/pull/17048)) by @2SinceJuly
- feat(web-shell): make the local files bridge easier to discover and use ([#17198](https://github.com/QwenLM/qwen-code/pull/17198)) by @wenshao
- feat: make each surface's resolved agent-loop path visible ([#17256](https://github.com/QwenLM/qwen-code/pull/17256)) by @wenshao
- feat(managed-agent): enable external workspace execution ([#15684](https://github.com/QwenLM/qwen-code/pull/15684)) by @yiliang114
- feat(verify-pr): require a boundary declaration and own the watch state ([#17185](https://github.com/QwenLM/qwen-code/pull/17185)) by @wenshao
- feat(channels): digest bound-PR comment backlog after a downtime gap ([#17199](https://github.com/QwenLM/qwen-code/pull/17199)) by @wenshao
- feat(core): run headless sessions and the Ink TUI on the shared agent loop by default ([#17362](https://github.com/QwenLM/qwen-code/pull/17362)) by @qqqys
- feat(serve): localize the daemon PR-watch prompt and pin the recorded follow-ups ([#17352](https://github.com/QwenLM/qwen-code/pull/17352)) by @wenshao
- feat(managed-agent): publish the managed /loop skill variant by profile ([#17272](https://github.com/QwenLM/qwen-code/pull/17272)) by @wenshao
- feat(web-shell): Complete the Russian Web Shell UI catalog ([#17414](https://github.com/QwenLM/qwen-code/pull/17414)) by @callmeYe
- feat(core): retain the latest complete operation during compaction ([#17141](https://github.com/QwenLM/qwen-code/pull/17141)) by @callmeYe
- feat(web-shell): add active highlight styling to selected reasoning e… ([#16622](https://github.com/QwenLM/qwen-code/pull/16622)) by @pranavshankar1221
- feat(core): run subagents on the shared agent loop by default ([#17420](https://github.com/QwenLM/qwen-code/pull/17420)) by @qqqys
- feat(web-shell): show per-turn hook statistics ([#17327](https://github.com/QwenLM/qwen-code/pull/17327)) by @callmeYe
- feat(acp): let sessions a program drives take accepted cross-session messages ([#17436](https://github.com/QwenLM/qwen-code/pull/17436)) by @qqqys
- feat(mods): let Mods rewrite expanded thinking in the terminals ([#17443](https://github.com/QwenLM/qwen-code/pull/17443)) by @qqqys
- feat(core): slim edit output and add authorized stale recovery ([#17255](https://github.com/QwenLM/qwen-code/pull/17255)) by @callmeYe
- feat(web-shell): surface a session-artifact header entry before the replay on cold open ([#17397](https://github.com/QwenLM/qwen-code/pull/17397)) by @wenshao
- feat(generated-ui): store card view state in the session transcript ([#17448](https://github.com/QwenLM/qwen-code/pull/17448)) by @qqqys
- feat(web-shell): make Goal state and session navigation readable at a glance ([#17349](https://github.com/QwenLM/qwen-code/pull/17349)) by @wenshao
- feat(web-shell): resolve dirty-tree checkout refusals in the branch picker ([#17364](https://github.com/QwenLM/qwen-code/pull/17364)) by @wenshao
- feat(serve): run a create_sub_session sub-session in another workspace ([#17454](https://github.com/QwenLM/qwen-code/pull/17454)) by @qqqys
- feat(web-shell): add delivery-unit change overview for sessions ([#17378](https://github.com/QwenLM/qwen-code/pull/17378)) by @wenshao
- feat(channels): add Aone Code channel plugin with polling event-stream inbound ([#17442](https://github.com/QwenLM/qwen-code/pull/17442)) by @wenshao
- feat(web-shell): Codex-style managed worktree sessions (default-on, detached, branch & handoff) ([#17228](https://github.com/QwenLM/qwen-code/pull/17228)) by @wenshao
- feat(serve): let create_sub_session follow up a sub-session the caller created ([#17484](https://github.com/QwenLM/qwen-code/pull/17484)) by @qqqys
- feat(web-shell): add activity date session grouping ([#17447](https://github.com/QwenLM/qwen-code/pull/17447)) by @callmeYe
- feat: add qwen4-max support with native effort ladder, cache TTL, and image detail ([#17513](https://github.com/QwenLM/qwen-code/pull/17513)) by @doudouOUC

### Bug Fixes

- fix(managed-agent): settle stale public-child recovery as terminal instead of retrying forever ([#17273](https://github.com/QwenLM/qwen-code/pull/17273)) by @yiliang114
- fix(sdk): route Skill mutations through native REST ([#17111](https://github.com/QwenLM/qwen-code/pull/17111)) by @Afloat16
- fix(sdk): keep the capabilities deadline through body reads ([#17103](https://github.com/QwenLM/qwen-code/pull/17103)) by @Afloat16
- fix(sdk): release settled WebSocket probe deadlines ([#17107](https://github.com/QwenLM/qwen-code/pull/17107)) by @Afloat16
- fix(cli): preserve surfaced task selection across reordering ([#17126](https://github.com/QwenLM/qwen-code/pull/17126)) by @Afloat16
- fix(cli): ignore unchanged JSON todo snapshots ([#17122](https://github.com/QwenLM/qwen-code/pull/17122)) by @Afloat16
- fix(web-shell): preserve withheld memory content as read-only ([#17109](https://github.com/QwenLM/qwen-code/pull/17109)) by @Afloat16
- fix(review): propagate explicit run deadlines to the child ([#17118](https://github.com/QwenLM/qwen-code/pull/17118)) by @Afloat16
- fix(cli): preserve the auth dialog exit confirmation window ([#17125](https://github.com/QwenLM/qwen-code/pull/17125)) by @Afloat16
- fix(cli): show disabled multiselect rows by selection state ([#17124](https://github.com/QwenLM/qwen-code/pull/17124)) by @Afloat16
- fix(core): retain buffered shell output after process exit ([#17112](https://github.com/QwenLM/qwen-code/pull/17112)) by @Afloat16
- fix(web-shell): clear Goals multi-workspace post-merge review items ([#17291](https://github.com/QwenLM/qwen-code/pull/17291)) by @wenshao
- fix(core): recheck transcript file type on the opened fd (TOCTOU hardening) ([#17301](https://github.com/QwenLM/qwen-code/pull/17301)) by @wenshao
- fix(web-shell): finalize voice dictation before sending a message ([#17331](https://github.com/QwenLM/qwen-code/pull/17331)) by @wenshao
- fix(acp): skip unusable transcript refreshes during active loads ([#17106](https://github.com/QwenLM/qwen-code/pull/17106)) by @Afloat16
- fix(cli): hide root-restricted workspace settings ([#17121](https://github.com/QwenLM/qwen-code/pull/17121)) by @Afloat16
- fix(sdk): refresh replay capability after transport reconnects ([#17110](https://github.com/QwenLM/qwen-code/pull/17110)) by @Afloat16
- fix(cli): record cancelled tool results on the headless shared loop ([#17353](https://github.com/QwenLM/qwen-code/pull/17353)) by @qqqys
- fix(cli): preserve hosted turn history after interruption ([#15959](https://github.com/QwenLM/qwen-code/pull/15959)) by @water-in-stone
- fix(channels): land the managed-channel review follow-ups (#17173) ([#17351](https://github.com/QwenLM/qwen-code/pull/17351)) by @wenshao
- fix(cli): preserve full commands in execution confirmations ([#17361](https://github.com/QwenLM/qwen-code/pull/17361)) by @chiga0
- fix(channels): evict refused DingTalk webhooks and surface admission diagnostics ([#17396](https://github.com/QwenLM/qwen-code/pull/17396)) by @wenshao
- fix(cli): narrow /effort to the tiers the active model supports ([#17425](https://github.com/QwenLM/qwen-code/pull/17425)) by @DragonnZhang
- fix(verify-pr): close the context trust refusal's root-tmp and hard-link bypasses ([#17375](https://github.com/QwenLM/qwen-code/pull/17375)) by @wenshao
- fix(cli): let a rerun join go on with the credential its first run saved ([#16831](https://github.com/QwenLM/qwen-code/pull/16831)) by @wenshao
- fix(channels): guard every managed DingTalk persist by its side-effect position ([#17432](https://github.com/QwenLM/qwen-code/pull/17432)) by @wenshao
- fix(managed-agent): Wait for workspace preparation before turn admission ([#17457](https://github.com/QwenLM/qwen-code/pull/17457)) by @doudouOUC
- fix(serve): accept hook description and statusMessage during extension preparation ([#17439](https://github.com/QwenLM/qwen-code/pull/17439)) by @doudouOUC
- fix(managed-agent): read the content-only approval selection at public-child version 1 ([#17428](https://github.com/QwenLM/qwen-code/pull/17428)) by @yiliang114
- fix(managed-agent): resolve external execution actor through child session lineage ([#17412](https://github.com/QwenLM/qwen-code/pull/17412)) by @yiliang114
- fix(managed-agent): resolve shared workspace owners for extension control ([#17480](https://github.com/QwenLM/qwen-code/pull/17480)) by @doudouOUC
- fix(managed-agent): recover macOS workspaces after reboot ([#17466](https://github.com/QwenLM/qwen-code/pull/17466)) by @doudouOUC
- fix(managed-agent): recover timed-out hosted journal commits ([#17459](https://github.com/QwenLM/qwen-code/pull/17459)) by @doudouOUC
- fix(web-shell): classify the directory-collision checkout refusal and close post-merge gaps ([#17476](https://github.com/QwenLM/qwen-code/pull/17476)) by @wenshao
- fix(channels): retire the DingTalk SDK heartbeat interval on socket close ([#17462](https://github.com/QwenLM/qwen-code/pull/17462)) by @wenshao
- fix(cli): avoid watching unrelated settings entries ([#17475](https://github.com/QwenLM/qwen-code/pull/17475)) by @chiga0
- fix(channels): give each DingTalk reply chunk its own send budget ([#17504](https://github.com/QwenLM/qwen-code/pull/17504)) by @wenshao
- fix(deps): patch express-rate-limit window accounting ([#17525](https://github.com/QwenLM/qwen-code/pull/17525)) by @chiga0
- fix(core): accept native ripgrep regex syntax in Grep ([#14315](https://github.com/QwenLM/qwen-code/pull/14315)) by @dakjdakd

### Documentation

- docs(design): add the agent loop's capability ledger and rollout rules ([#17285](https://github.com/QwenLM/qwen-code/pull/17285)) by @qqqys
- docs(serve): document the two ignored watch knobs and calibrate the bound-delivery designs ([#17309](https://github.com/QwenLM/qwen-code/pull/17309)) by @wenshao
- docs(design): record speculation PR 6b decision to keep its own tool step ([#17324](https://github.com/QwenLM/qwen-code/pull/17324)) by @wenshao
- docs(design): close the owed claims #17352 landed and the stale role-guidance status ([#17395](https://github.com/QwenLM/qwen-code/pull/17395)) by @wenshao
- docs(users): add Web Shell voice, Live, and image generation guides ([#17347](https://github.com/QwenLM/qwen-code/pull/17347)) by @wenshao
- docs(design): record the landings the channel design statuses still hide ([#17409](https://github.com/QwenLM/qwen-code/pull/17409)) by @wenshao
- docs(design): bring the rollout-switch design up to date with main ([#17410](https://github.com/QwenLM/qwen-code/pull/17410)) by @qqqys
- docs(design): reconcile the unified-loop end state with the keep-its-own decisions ([#17416](https://github.com/QwenLM/qwen-code/pull/17416)) by @wenshao
- docs(design): propose a unified chat surface for daemon and managed sessions ([#17460](https://github.com/QwenLM/qwen-code/pull/17460)) by @wenshao

### Internal Changes

- chore(deps): update vulnerable dependencies ([#13169](https://github.com/QwenLM/qwen-code/pull/13169)) by @LubabaKhalid
- test(cli): cover the shared batch host inputs on cron and notification turns ([#17330](https://github.com/QwenLM/qwen-code/pull/17330)) by @yiliang114
- test(web-shell): stabilize global turn navigation scroll-follow E2E (#17252) ([#17259](https://github.com/QwenLM/qwen-code/pull/17259)) by @qwen-code-dev-bot
- chore(deps): clear remaining CVE audit findings (diff DoS, uuid bounds check) ([#10985](https://github.com/QwenLM/qwen-code/pull/10985)) by @now-ing
- test(web-shell): stabilize the workflow-mode policy read test (#17348) ([#17355](https://github.com/QwenLM/qwen-code/pull/17355)) by @qwen-code-dev-bot
- test(core): deflake the unawaited-timer code mode test ([#17345](https://github.com/QwenLM/qwen-code/pull/17345)) by @qwen-code-dev-bot
- test(integration): settle the ACP parity debug log before stopping the agent ([#17365](https://github.com/QwenLM/qwen-code/pull/17365)) by @qwen-code-dev-bot
- test(ci): stabilize the macOS nightly CI legs ([#17219](https://github.com/QwenLM/qwen-code/pull/17219)) by @wenshao
- ci(release): retry the Docker integration test suites once ([#16862](https://github.com/QwenLM/qwen-code/pull/16862)) by @qwen-code-dev-bot
- test(core): cover speculation's partial-batch boundary hand-back on both arms ([#17399](https://github.com/QwenLM/qwen-code/pull/17399)) by @wenshao
- test(cli): wait for the applied skills selection before navigating ([#17413](https://github.com/QwenLM/qwen-code/pull/17413)) by @qwen-code-dev-bot
- test(channels): pin the managed DingTalk lifecycle, settings, and export surfaces ([#17417](https://github.com/QwenLM/qwen-code/pull/17417)) by @wenshao
- test(channels): pin the managed DingTalk stream wiring and webhook send budget ([#17426](https://github.com/QwenLM/qwen-code/pull/17426)) by @wenshao
- refactor(managed-agent): archive Kubernetes runtime and resident CSI backend ([#17415](https://github.com/QwenLM/qwen-code/pull/17415)) by @doudouOUC
- test(generated-ui): read comma-joined amounts and drop an unobservable check ([#17430](https://github.com/QwenLM/qwen-code/pull/17430)) by @qqqys
- test(core): characterize leader delivery and terminal state per host class ([#14855](https://github.com/QwenLM/qwen-code/pull/14855)) by @tonydzi
- test(managed-agent): add a real model-turn E2E for the /loop skill ([#17433](https://github.com/QwenLM/qwen-code/pull/17433)) by @wenshao
- test(managed-runtime): pin the deferred acquire against close ([#16492](https://github.com/QwenLM/qwen-code/pull/16492)) by @2SinceJuly
- test(serve): pin description and statusMessage acceptance on managed hooks ([#17393](https://github.com/QwenLM/qwen-code/pull/17393)) by @MohammedAlkindi
- test(web-shell): deflake large source file smoke test ([#17373](https://github.com/QwenLM/qwen-code/pull/17373)) by @qwen-code-dev-bot
- test(managed-agent): restore the round-3 /loop E2E fixes and wire the scenario into CI ([#17470](https://github.com/QwenLM/qwen-code/pull/17470)) by @wenshao
- ci(review): opt the review's own commands into the execution sandbox ([#17282](https://github.com/QwenLM/qwen-code/pull/17282)) by @yiliang114
- refactor(managed-agent): unify coordinator dispatch deduplication and lease renewal ([#17508](https://github.com/QwenLM/qwen-code/pull/17508)) by @doudouOUC
- refactor(managed-agent): replace regex DDL catalog parsing with a tokenizing parser ([#17509](https://github.com/QwenLM/qwen-code/pull/17509)) by @doudouOUC
- chore(managed-agent): deduplicate verification Failsafe profiles and land #17139 follow-ups ([#17507](https://github.com/QwenLM/qwen-code/pull/17507)) by @doudouOUC
- test(web-shell): restore the label-only adoption witness and reconcile the classifier's documented reach ([#17503](https://github.com/QwenLM/qwen-code/pull/17503)) by @wenshao
- test(cli): bound managed terminal watchdogs and teardown by host budget ([#17268](https://github.com/QwenLM/qwen-code/pull/17268)) by @qwen-code-dev-bot
- refactor(managed-agent): require canonical Broker URLs at the preflight and Harness entry ([#17510](https://github.com/QwenLM/qwen-code/pull/17510)) by @doudouOUC
- test(integration): settle the second turn before reloading the generated-UI session ([#17537](https://github.com/QwenLM/qwen-code/pull/17537)) by @qwen-code-dev-bot

## New Contributors

- @LubabaKhalid made their first contribution in [#13169](https://github.com/QwenLM/qwen-code/pull/13169)
- @pranavshankar1221 made their first contribution in [#16622](https://github.com/QwenLM/qwen-code/pull/16622)

**Full Changelog**: https://github.com/QwenLM/qwen-code/compare/v0.27.4...v0.28.0
← Older: v0.27.4-nightly.20261116.eb3a599ef3 (2026-11-16)Newer: v0.28.1-preview.0 (2026-11-17)
Weekly digest: every breaking change, retirement and price change across 23 coding agents, one email.

Form not loading? Subscribe here.