Tools / Qwen Code / v0.27.4

Qwen Code v0.27.4 release notes & breaking changes

breakingsecuritymodel

Qwen Code v0.27.4 is a Alibaba release published 2026-11-15 with 97 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

- Managed Agent sessions can access granted local directories and open an interactive Managed Web Terminal without persisting credentials or paths. ([#16844](https://github.com/QwenLM/qwen-code/pull/16844), [#16866](https://github.com/QwenLM/qwen-code/pull/16866))
- Send outputs to Feishu and WeCom webhooks, with long managed-channel replies delivered as ordered segments instead of being truncated. ([#17012](https://github.com/QwenLM/qwen-code/pull/17012), [#17013](https://github.com/QwenLM/qwen-code/pull/17013), [#17223](https://github.com/QwenLM/qwen-code/pull/17223), [#17224](https://github.com/QwenLM/qwen-code/pull/17224), [#17002](https://github.com/QwenLM/qwen-code/pull/17002))
- Auto Mode now requires explicit authorization for risky actions, reviews classifier timeouts, avoids HTTP(S) URL filesystem lookups, and visibly marks Full Access. ([#16858](https://github.com/QwenLM/qwen-code/pull/16858), [#17127](https://github.com/QwenLM/qwen-code/pull/17127), [#17146](https://github.com/QwenLM/qwen-code/pull/17146), [#17100](https://github.com/QwenLM/qwen-code/pull/17100))
- Web Shell Goals page can merge goals from trusted workspaces, show workspace badges, and target a selected workspace when creating goals. ([#17230](https://github.com/QwenLM/qwen-code/pull/17230))
- Manage memory with restored CLI /memory controls, persistent workspace settings, and typed SDK APIs to preview, submit, inspect, cancel, and retry Codex memory imports. ([#17162](https://github.com/QwenLM/qwen-code/pull/17162), [#17213](https://github.com/QwenLM/qwen-code/pull/17213))
- Terminal tabs, web previews, local commands, and slash commands keep working over plain HTTP when native UUID support is unavailable. ([#17196](https://github.com/QwenLM/qwen-code/pull/17196))

## Breaking Changes

No known breaking changes.

## Complete Change List

### Features

- Managed Agent sessions can use the browser user's local filesystem through temporary directory grants, without persisting credentials or paths. ([#16844](https://github.com/QwenLM/qwen-code/pull/16844)) by @wenshao
- Long managed-channel replies are split into ordered delivery segments instead of being truncated, with a visible truncation notice if the segment limit is exceeded. ([#17002](https://github.com/QwenLM/qwen-code/pull/17002)) by @wenshao
- With experimental.unifiedAgentLoop enabled, each AgentCore round runs model and tool batch in one runAgentLoop call, optionally on the event-bus kernel. ([#17059](https://github.com/QwenLM/qwen-code/pull/17059)) by @qqqys
- feat(cli): run each OpenTUI round in one runAgentLoop call ([#17104](https://github.com/QwenLM/qwen-code/pull/17104)) by @qqqys
- feat(channels): add Feishu custom bot webhook delivery ([#17012](https://github.com/QwenLM/qwen-code/pull/17012)) by @jkxiao
- Auto Mode now lets the session model review a fast-classifier timeout once when identifiers differ, allowing or blocking the action under existing policy. ([#17146](https://github.com/QwenLM/qwen-code/pull/17146)) by @callmeYe
- Exposes a typed workspace-bound SDK API for previewing, submitting, inspecting, canceling, and retrying Codex memory imports. ([#17162](https://github.com/QwenLM/qwen-code/pull/17162)) by @callmeYe
- Runs each Ink TUI round in one runAgentLoop call when experimental.unifiedAgentLoop is enabled, aligning TUI loop execution. ([#17160](https://github.com/QwenLM/qwen-code/pull/17160)) by @qqqys
- Adds per-turn Hook execution records, read-only Hook pages, SDK access, summaries, and signed snapshots for session history. ([#17015](https://github.com/QwenLM/qwen-code/pull/17015)) by @callmeYe
- Makes cancelled runs publish loop.cancelled and finish within a 10-second bound, preserving completed batch results when possible. ([#17145](https://github.com/QwenLM/qwen-code/pull/17145)) by @qqqys
- Adds a send-only Feishu Webhook connection method with secret URL, optional signing secret, channel control, and scheduled group delivery. ([#17013](https://github.com/QwenLM/qwen-code/pull/17013)) by @jkxiao
- Adds nightly E2E loop rollout legs for kernel and OpenTUI, including forced parity arms, while push runs keep their current legs. ([#17050](https://github.com/QwenLM/qwen-code/pull/17050)) by @wenshao
- Enables the serve daemon's PR-state watch to observe and deliver PR updates promptly by provisioning its dependencies and adding resume delivery. ([#17032](https://github.com/QwenLM/qwen-code/pull/17032)) by @wenshao
- Improves web-shell session search with a dedicated centered dialog, clearer results, keyboard selection, loading and retry states, and responsive layouts. ([#16779](https://github.com/QwenLM/qwen-code/pull/16779)) by @deggs7
- Adds an interactive Managed Web Terminal to Managed Agent sessions, using a short-lived ticket for authenticated WebSocket access and keeping the runtime active while attached. ([#16866](https://github.com/QwenLM/qwen-code/pull/16866)) by @wenshao
- feat(managed-agent): fixed-interval /loop schedules for Managed Sessions ([#17088](https://github.com/QwenLM/qwen-code/pull/17088)) by @wenshao
- With experimental.unifiedAgentLoop enabled, each ACP round runs in one runAgentLoop call; experimental.eventBusAgentLoop uses the event-bus kernel. ([#17181](https://github.com/QwenLM/qwen-code/pull/17181)) by @qqqys
- When experimental.unifiedAgentLoop is enabled, speculation model rounds now use the shared agent loop; otherwise behavior is unchanged. ([#17029](https://github.com/QwenLM/qwen-code/pull/17029)) by @wenshao
- Experimental loop switches now resolve per surface and can be forced on or off by environment overrides such as 1, 0, or headless,acp=0. ([#17055](https://github.com/QwenLM/qwen-code/pull/17055)) by @wenshao
- Adds a send-only WeCom group webhook Channel that can deliver Notify outputs, prompt results, and supported scheduled-task results to a configured WeCom group. ([#17223](https://github.com/QwenLM/qwen-code/pull/17223)) by @jkxiao
- Adds a send-only WeCom Webhook connection method with group URL configuration, channel control, and scheduled delivery selection. ([#17224](https://github.com/QwenLM/qwen-code/pull/17224)) by @jkxiao
- Makes PR watch operator-owned by adding the operator-only serve.sessionPrWatch setting and retiring GitHub channel state-transition delivery. ([#17214](https://github.com/QwenLM/qwen-code/pull/17214)) by @wenshao
- Restored native /memory controls and User/Project file or managed-folder actions in the CLI, with workspace settings persisting across restarts. ([#17213](https://github.com/QwenLM/qwen-code/pull/17213)) by @chiga0
- Adds multi-workspace support to the Web Shell Goals page, merging goals from trusted workspaces, adding workspace badges, and letting new goals target a selected workspace. ([#17230](https://github.com/QwenLM/qwen-code/pull/17230)) by @wenshao
- Adds experimental.unifiedAgentLoop hosted surface so Hosted Harness model rounds can use the shared agent loop while the legacy path remains default. ([#17264](https://github.com/QwenLM/qwen-code/pull/17264)) by @wenshao

### Bug Fixes

- AUTO permission checks now judge explicit user authorization, especially for destructive, discarding, shared-history, or bulk actions, rather than mere relevance. ([#16858](https://github.com/QwenLM/qwen-code/pull/16858)) by @yanyukang2003-ux
- When Full Access (yolo) is active, the web shell composer mode trigger now shows its icon and label in the destructive color, keeping the risky mode visibly marked. ([#17100](https://github.com/QwenLM/qwen-code/pull/17100)) by @callmeYe
- Synchronous managed-session rename settlement no longer waits on the stale tenant guard, allowing concurrent first renames to settle independently within or across tenants. ([#17074](https://github.com/QwenLM/qwen-code/pull/17074)) by @doudouOUC
- fix(web-shell): preserve observer activity during recovery ([#17098](https://github.com/QwenLM/qwen-code/pull/17098)) by @callmeYe
- fix(web-shell): show timeline on narrower screens ([#16986](https://github.com/QwenLM/qwen-code/pull/16986)) by @callmeYe
- fix(deps): 修复 MCP、DOMPurify 和压缩请求解析漏洞 ([#17105](https://github.com/QwenLM/qwen-code/pull/17105)) by @chiga0
- fix(managed-agent): stop the legacy CSI guard from gap-locking other Sessions ([#17069](https://github.com/QwenLM/qwen-code/pull/17069)) by @doudouOUC
- Updated qs, fast-uri, ip-address, Hono, and Node adapter security dependencies across workspace and Mobile MCP trees. ([#17134](https://github.com/QwenLM/qwen-code/pull/17134)) by @chiga0
- AUTO mode no longer tries filesystem resolution for HTTP(S) URLs, preventing Windows network lookup delays and session termination before requests start. ([#17127](https://github.com/QwenLM/qwen-code/pull/17127)) by @callmeYe
- Native Anthropic conversations now omit unsigned thinking blocks from completed history while keeping visible answers, signed thinking, and completed tool exchanges. ([#17031](https://github.com/QwenLM/qwen-code/pull/17031)) by @callmeYe
- Updated fast-uri and ip-address in the independent Live Host and CUA Electron installations to secure versions, closing remaining URI and IPv6 advisories. ([#17147](https://github.com/QwenLM/qwen-code/pull/17147)) by @chiga0
- Goal status messages now show a compact one-line objective preview with Expand details, keeping long objectives from filling the chat. ([#17133](https://github.com/QwenLM/qwen-code/pull/17133)) by @callmeYe
- An unconfirmed send from another conversation no longer locks the composer; it remains editable, but submission is blocked until retry or discard. ([#17140](https://github.com/QwenLM/qwen-code/pull/17140)) by @yiliang114
- Edit now recovers matches between non-ASCII text and Unicode escape spellings when exact matching fails, preserving the file's original representation. ([#17096](https://github.com/QwenLM/qwen-code/pull/17096)) by @callmeYe
- Raises the document export renderer size budget to 2,136,000 bytes so current builds pass export validation. ([#17153](https://github.com/QwenLM/qwen-code/pull/17153)) by @chiga0
- Upgrades glob and minimatch together to secure versions, improving matching reliability and preventing search regressions. ([#17149](https://github.com/QwenLM/qwen-code/pull/17149)) by @chiga0
- Hover timestamps no longer cover message controls in the web shell, keeping buttons such as Goal details readable. ([#17168](https://github.com/QwenLM/qwen-code/pull/17168)) by @callmeYe
- Removes obsolete ExcelJS archive dependency paths and updates unzipper to 0.12.5 for cleaner dependency resolution. ([#17163](https://github.com/QwenLM/qwen-code/pull/17163)) by @chiga0
- Upgrades minimatch and glob dependencies in independent Electron build trees to address security advisories. ([#17165](https://github.com/QwenLM/qwen-code/pull/17165)) by @chiga0
- fix(external-context): deflake malformed proxy env test (#17176) ([#17179](https://github.com/QwenLM/qwen-code/pull/17179)) by @qwen-code-dev-bot
- Prevents table toolbar overlap in narrow web-shell views by compacting controls, shortening summaries, and preserving row counts. ([#17020](https://github.com/QwenLM/qwen-code/pull/17020)) by @Lilian0122
- Clears non-sensitive OpenTUI extension-setting drafts when Ctrl+C is pressed, fixing text accumulation in ordinary setting fields. ([#15951](https://github.com/QwenLM/qwen-code/pull/15951)) by @chiga0
- Prepares file watchers before final interactive exit so /quit can complete cleanly instead of leaving the CLI process alive. ([#17148](https://github.com/QwenLM/qwen-code/pull/17148)) by @chiga0
- Fixes Unicode Edit recovery so it no longer damages syntax by matching inside backslash, hexadecimal, or Unicode escape sequences. ([#17152](https://github.com/QwenLM/qwen-code/pull/17152)) by @callmeYe
- fix(ci): route classify_pr by association and admit dispatch to the fleet ([#17178](https://github.com/QwenLM/qwen-code/pull/17178)) by @yiliang114
- Terminal tabs, web previews, local and slash commands, and related shell flows now keep working over plain HTTP by generating IDs when native UUID support is unavailable. ([#17196](https://github.com/QwenLM/qwen-code/pull/17196)) by @wenshao
- Managed Sessions /loop schedules now recover more efficiently and only expose the cron tools declared for the session profile. ([#17197](https://github.com/QwenLM/qwen-code/pull/17197)) by @wenshao
- Fixes the agent tab context usage meter to use the agent's own model window instead of the main session window. ([#13614](https://github.com/QwenLM/qwen-code/pull/13614)) by @Shizoqua
- Fixed the CLI feedback prompt to preserve typed input and made newly registered workflow commands appear immediately in the model-facing catalog. ([#17233](https://github.com/QwenLM/qwen-code/pull/17233)) by @qqqys
- Fixed managed-agent configuration reads to accept PUBLISHED as a settled state for the relevant Agent resources. ([#17207](https://github.com/QwenLM/qwen-code/pull/17207)) by @yiliang114
- Fixed the memory-manager integration suite so it can choose an output root without requiring a special environment variable. ([#17265](https://github.com/QwenLM/qwen-code/pull/17265)) by @wenshao
- Returns correlated error responses for Java SDK control callback failures, including request ID, subtype, and error value. ([#17113](https://github.com/QwenLM/qwen-code/pull/17113)) by @Afloat16
- Fixes release validation failures in interactive memory-manager tests by ensuring fake editor and opener scripts run correctly. ([#17294](https://github.com/QwenLM/qwen-code/pull/17294)) by @qwen-code-dev-bot
- Fixes managed agent child sessions to use inherited extension configuration at committed state and report failures before Runtime acquisition. ([#17281](https://github.com/QwenLM/qwen-code/pull/17281)) by @yiliang114

### Performance

- Improves CLI performance in ACP by skipping interactive extension watching, reducing startup contention. ([#17177](https://github.com/QwenLM/qwen-code/pull/17177)) by @callmeYe
- Speeds up the first native ACP session by overlapping initialization work and deferring unneeded command and dependency loading. ([#17209](https://github.com/QwenLM/qwen-code/pull/17209)) by @callmeYe
- Improves Java transcript performance by sharing immutable event snapshots instead of duplicating parsed page data. ([#17115](https://github.com/QwenLM/qwen-code/pull/17115)) by @Afloat16
- Speeds up the Web Shell Daemon Status Usage dashboard by caching per-transcript usage summaries and rescanning only changed transcripts. ([#17239](https://github.com/QwenLM/qwen-code/pull/17239)) by @wenshao

### Documentation

- Adds bilingual Managed Agent design documentation inventorying lifecycle claim mirrors and tenant guards without changing runtime behavior. ([#17073](https://github.com/QwenLM/qwen-code/pull/17073)) by @doudouOUC
- Documents streaming tool dispatch decisions in English and Chinese, with opt-in implementation deferred to a follow-up. ([#17171](https://github.com/QwenLM/qwen-code/pull/17171)) by @qqqys
- README Capabilities now presents a grouped overview of Qwen Code features, with links to user docs and labels for experimental and opt-in features. ([#17226](https://github.com/QwenLM/qwen-code/pull/17226)) by @wenshao
- Added design documentation evaluating the Hosted Harness agent-loop migration and its separation into journaled tool batch and model round paths. ([#17248](https://github.com/QwenLM/qwen-code/pull/17248)) by @wenshao
- Updated README availability claims to distinguish default features from experimental or opt-in features and expanded listed chat channels. ([#17263](https://github.com/QwenLM/qwen-code/pull/17263)) by @wenshao
- Updates README capability labels to avoid repeated pitch text and links /resume to the commands reference. ([#17277](https://github.com/QwenLM/qwen-code/pull/17277)) by @wenshao
- Updates PR watch design documents to record shipped status, settled review questions, and the removal of the enablement switch. ([#17250](https://github.com/QwenLM/qwen-code/pull/17250)) by @wenshao

### Internal Changes

- Web shell test now opens Session Overview from the footer entry instead of the More menu, matching single-column layouts. ([#17067](https://github.com/QwenLM/qwen-code/pull/17067)) by @qwen-code-dev-bot
- Managed Agent schema initialization now uses ManagedSchema.initialize directly after removing the redundant JdbcRuntimeBrokerSchema alias. ([#17075](https://github.com/QwenLM/qwen-code/pull/17075)) by @doudouOUC
- refactor(managed-agent): drop the catalog probe and repeated root lock from binding control ([#17072](https://github.com/QwenLM/qwen-code/pull/17072)) by @doudouOUC
- test(cli): compare the kernel off and on in every headless parity scenario ([#17130](https://github.com/QwenLM/qwen-code/pull/17130)) by @qqqys
- Improves ACP turn handling by replacing host-ownership flags with explicit turn inputs and caller-owned interaction spans. ([#17070](https://github.com/QwenLM/qwen-code/pull/17070)) by @yiliang114
- Adds an embedding regression test and removes obsolete release repair steps to preserve Managed Agent scheduler isolation. ([#17139](https://github.com/QwenLM/qwen-code/pull/17139)) by @doudouOUC
- Stabilizes the TUI background notification parity test by normalizing launch size fields and PID-width variations. ([#17114](https://github.com/QwenLM/qwen-code/pull/17114)) by @qwen-code-dev-bot
- Stabilizes headless and TUI agent-loop parity tests to reduce flaky CI failures without weakening assertions. ([#17172](https://github.com/QwenLM/qwen-code/pull/17172)) by @yiliang114
- Moves the mid-width toolbar compaction test into the intended container-width band; no product behavior changes. ([#17187](https://github.com/QwenLM/qwen-code/pull/17187)) by @wenshao
- Stabilizes the managed runtime terminal replay test by attaching the output listener before WebSocket reconnection; no product behavior changes. ([#17194](https://github.com/QwenLM/qwen-code/pull/17194)) by @qwen-code-dev-bot
- Adds a CI lane that verifies shared-loop and event-bus agent execution across Managed headless, AgentCore, and ACP entry points. ([#17180](https://github.com/QwenLM/qwen-code/pull/17180)) by @yiliang114
- CI stability improves for large transcript export tests on shared ARM64 runners without weakening regression detection on GitHub-hosted CI or local runs. ([#17205](https://github.com/QwenLM/qwen-code/pull/17205)) by @qwen-code-dev-bot
- Adds tests verifying Managed Sessions automation handling of null creator keys and tenant isolation, with no production behavior change. ([#17235](https://github.com/QwenLM/qwen-code/pull/17235)) by @wenshao
- test(external-context): clear lowercase proxy vars in the fail-fast pack test ([#17240](https://github.com/QwenLM/qwen-code/pull/17240)) by @wenshao
- Internal test change: deflakes the queued-input parity case by treating process-derived shell result size telemetry as volatile. ([#17220](https://github.com/QwenLM/qwen-code/pull/17220)) by @qwen-code-dev-bot
- Internal test change: strengthens CI coverage for tree-ref encoding guards and corrects misleading test comments. ([#17191](https://github.com/QwenLM/qwen-code/pull/17191)) by @wenshao
- Internal test change: makes the external-context fail-fast pack test self-contained on proxied runners and adjusts its subprocess timeout. ([#17246](https://github.com/QwenLM/qwen-code/pull/17246)) by @qwen-code-dev-bot
- Improved web-shell retention tests so wheel events target the transcript and virtualized card restore checks run reliably. ([#17242](https://github.com/QwenLM/qwen-code/pull/17242)) by @wenshao
- Improved ACP integration tests by counting loop rounds with a dedicated marker for each runAgentLoop call. ([#17227](https://github.com/QwenLM/qwen-code/pull/17227)) by @qqqys
- Removes an unused per-call client ID override from Web Shell QR-login wiring without changing login behavior. ([#16999](https://github.com/QwenLM/qwen-code/pull/16999)) by @wen
← Older: v0.27.2-nightly.20261113.6f754be654 (2026-11-14)Newer: v0.27.3-nightly.20261115.6e3b2dcc19 (2026-11-15)
Weekly digest: every breaking change, retirement and price change across 23 coding agents, one email.

Form not loading? Subscribe here.