Qwen Code

<!-- Release notes generated using configuration in .github/release.yml at b74592785bff038a1e3b5c533d20ae49e549ab2a -->

What's Changed

• ci(desktop): exercise desktop packaging on a schedule by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/11519

• fix(bridge): keep pending permission/question across refreshed session loads by @wenshao in https://github.com/QwenLM/qwen-code/pull/11468

• test(integration): deflake the /compress E2E event budget (#11088) by @qwen-code-dev-bot in https://github.com/QwenLM/qwen-code/pull/11094

• fix(export): restore document runtime budget headroom by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/11372

• fix(core): preserve Responses stream error details by @tanzhenxin in https://github.com/QwenLM/qwen-code/pull/11523

• fix(core): release node-pty's conout worker after every PTY on Windows by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/11313

• fix(web-shell): make the workspace provider guard self-diagnosing and reload on root retry by @wenshao in https://github.com/QwenLM/qwen-code/pull/11421

Preview build. This is a prerelease: the desktop-latest updater feed still points at 0.2.2, so existing installations are not moved onto it. Install it deliberately if you want to try it.

Downloads

The .app.tar.gz archives are updater payloads, not downloads. SHA256SUMS.txt covers every asset.

Windows shows a SmartScreen warning

Expected, and not specific to this build: Qwen Code Desktop installers have never carried an Authenticode signature, because no Windows code signing certificate is configured for this repository. Windows will show "Windows protected your PC" on first install — choose More info, then Run anyway. Verify the download against SHA256SUMS.txt if you want to confirm what you are running.

Qwen CUA Driver prebuilt binaries (vendored under packages/cua-driver).

• macOS: codesigned + notarized universal binary + QwenCuaDriver.app
• Linux: unsigned (x86_64 + arm64, glibc 2.31 floor)
• Windows: unsigned UIAccess worker + native SDK payload (x86_64 + arm64); deployments must sign and trust the worker on the target machine before UIAccess activation
• Node.js: the same workflow publishes @qwen-code/cua-sdk after a clean install against these release assets and publishes the independently versioned @qwen-code/node-repl-mcp package after its MCP protocol smoke tests

Enable relative coordinates: CUA_DRIVER_RS_COORDINATE_SPACE=1 (default 0 = off; optional CUA_DRIVER_RS_COORDINATE_SCALE=1000).

Enable textual MCP payload filtering for affected model API routes: MCP_MODEL_PAYLOAD_FILTER=1 (default off).

<!-- Release notes generated using configuration in .github/release.yml at 655baed68fbd5f284ab50e18bb1cb4bbf11613f9 -->

What's Changed

• feat(web-shell): show what a Goal has spent against the window it is allowed by @qqqys in https://github.com/QwenLM/qwen-code/pull/11254

Bundled CLI Version

This SDK release bundles CLI version: 0.23.2

Source: CLI built from source (same branch/ref as SDK)

---

Bundled CLI Version

This SDK release bundles CLI version: 0.23.1

Source: CLI built from source (same branch/ref as SDK)

---

Highlights

The bundled CLI 0.23.0 includes two fixes requested in #11022:

• Managed memory availability now respects the memory.enableManagedAutoMemory setting (#6941). Hosts that disable managed auto memory no longer see remember/dream requests admitted or managed-memory instructions in the system prompt; local patches can be removed.

• Size-triggered microcompaction now clears old tool results toward a low watermark (#8464), so it no longer rewrites another part of the conversation prefix on every subsequent turn. Provider prompt-cache reuse is preserved across long tool-heavy sessions, reducing cached-input cost.

Bundled CLI Version

This SDK release bundles CLI version: 0.23.0

Source: latest stable CLI from npm

---

Bundled CLI Version

This SDK release bundles CLI version: 0.19.10

Source: latest stable CLI from npm

---

Bundled CLI Version

This SDK release bundles CLI version: 0.15.3

Bundled CLI Version

This SDK release bundles CLI version: 0.23.1

Source: CLI built from source (same branch/ref as SDK)

---

Highlights

The bundled CLI 0.23.0 includes two fixes requested in #11022:

• Managed memory availability now respects the memory.enableManagedAutoMemory setting (#6941). Hosts that disable managed auto memory no longer see remember/dream requests admitted or managed-memory instructions in the system prompt; local patches can be removed.

• Size-triggered microcompaction now clears old tool results toward a low watermark (#8464), so it no longer rewrites another part of the conversation prefix on every subsequent turn. Provider prompt-cache reuse is preserved across long tool-heavy sessions, reducing cached-input cost.

Bundled CLI Version

This SDK release bundles CLI version: 0.23.0

Source: latest stable CLI from npm

---

Bundled CLI Version

This SDK release bundles CLI version: 0.19.10

Source: latest stable CLI from npm

---

Bundled CLI Version

This SDK release bundles CLI version: 0.15.3

Source: latest stable CLI from npm

---

Bundled CLI Version

This SDK release bundles CLI version: 0.13.1

Highlights

The bundled CLI 0.23.0 includes two fixes requested in #11022:

• Managed memory availability now respects the memory.enableManagedAutoMemory setting (#6941). Hosts that disable managed auto memory no longer see remember/dream requests admitted or managed-memory instructions in the system prompt; local patches can be removed.

• Size-triggered microcompaction now clears old tool results toward a low watermark (#8464), so it no longer rewrites another part of the conversation prefix on every subsequent turn. Provider prompt-cache reuse is preserved across long tool-heavy sessions, reducing cached-input cost.

Bundled CLI Version

This SDK release bundles CLI version: 0.23.0

Source: latest stable CLI from npm

---

Bundled CLI Version

This SDK release bundles CLI version: 0.19.10

Source: latest stable CLI from npm

---

Bundled CLI Version

This SDK release bundles CLI version: 0.15.3

Source: latest stable CLI from npm

---

Bundled CLI Version

This SDK release bundles CLI version: 0.13.1

Source: latest stable CLI from npm

---

Backfilled release for npm-published version 0.1.5 (original workflow failed before GitHub release creation).

Qwen CUA Driver prebuilt binaries (vendored under packages/cua-driver).

• macOS: codesigned + notarized universal binary + QwenCuaDriver.app
• Linux: unsigned (x86_64 + arm64, glibc 2.31 floor)
• Windows: unsigned UIAccess worker + native SDK payload (x86_64 + arm64); deployments must sign and trust the worker on the target machine before UIAccess activation
• Node.js: the same workflow publishes @qwen-code/cua-sdk after a clean install against these release assets and publishes the independently versioned @qwen-code/node-repl-mcp package after its MCP protocol smoke tests

Enable relative coordinates: CUA_DRIVER_RS_COORDINATE_SPACE=1 (default 0 = off; optional CUA_DRIVER_RS_COORDINATE_SCALE=1000).

Enable textual MCP payload filtering for affected model API routes: MCP_MODEL_PAYLOAD_FILTER=1 (default off).

<!-- Release notes generated using configuration in .github/release.yml at 5014a091ab6e6374ec5afe747f4a0cd421f62dc4 -->

What's Changed

• feat(channels): add BTW side questions to Channel conversations by @qqqys in https://github.com/QwenLM/qwen-code/pull/10713

<!-- Release notes generated using configuration in .github/release.yml at 867bb94a1f317e6fae04c2f5b78961b42e030311 -->

What's Changed

• fix(ci): make shared ECS Vitest concurrency tunable by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/10667
• feat(cli): OpenTUI migration batch 4 — dialogs, commands, and session-rewind by @chiga0 in https://github.com/QwenLM/qwen-code/pull/10383
• feat(goal): let the model propose a Goal the user approves in a dialog by @qqqys in https://github.com/QwenLM/qwen-code/pull/10171
• fix(web-shell): persist model reasoning preferences by @callmeYe in https://github.com/QwenLM/qwen-code/pull/10489
• feat(web-shell): Add standalone chats by @doudouOUC in https://github.com/QwenLM/qwen-code/pull/10514
• fix(ci): keep web-shell visuals comment on upload failure by @AaronZ345 in https://github.com/QwenLM/qwen-code/pull/10253
• fix(permissions): never let the AUTO-mode classifier auto-approve out-of-workspace writes; always fall back to manual approval by @PratikWayase in https://github.com/QwenLM/qwen-code/pull/8744
• fix(review): gate decided stops on a composed re-rule verdict by @wenshao in https://github.com/QwenLM/qwen-code/pull/10310

Qwen CUA Driver prebuilt binaries (vendored under packages/cua-driver).

• macOS: codesigned + notarized universal binary + QwenCuaDriver.app
• Linux: unsigned (x86_64 + arm64, glibc 2.31 floor)
• Windows: unsigned UIAccess worker + native SDK payload (x86_64 + arm64); deployments must sign and trust the worker on the target machine before UIAccess activation
• Node.js: the same workflow publishes @qwen-code/cua-sdk after a clean install against these release assets and publishes the independently versioned @qwen-code/node-repl-mcp package after its MCP protocol smoke tests

Enable relative coordinates: CUA_DRIVER_RS_COORDINATE_SPACE=1 (default 0 = off; optional CUA_DRIVER_RS_COORDINATE_SCALE=1000).

Enable textual MCP payload filtering for affected model API routes: MCP_MODEL_PAYLOAD_FILTER=1 (default off).

<!-- Release notes generated using configuration in .github/release.yml at 33a75a6b0a1a6b4b405caa6f1bf1f9f47f122e5c -->

What's Changed

• fix(core): decouple permissions.allow from tool registration via tools.eager (#10075) by @yiliang114 in https://github.com/QwenLM/qwen-code/pull/10098

Qwen CUA Driver prebuilt binaries (vendored under packages/cua-driver).

• macOS: codesigned + notarized universal binary + QwenCuaDriver.app
• Linux: unsigned (x86_64 + arm64, glibc 2.31 floor)
• Windows: unsigned (x86_64 + arm64)
• Node.js: the same workflow publishes @qwen-code/cua-sdk after a clean install against these release assets and publishes the independently versioned @qwen-code/node-repl-mcp package after its MCP protocol smoke tests

Enable relative coordinates: CUA_DRIVER_RS_COORDINATE_SPACE=1 (default 0 = off; optional CUA_DRIVER_RS_COORDINATE_SCALE=1000).

Enable textual MCP payload filtering for affected model API routes: MCP_MODEL_PAYLOAD_FILTER=1 (default off).

<!-- Release notes generated using configuration in .github/release.yml at b89ba11ede3f40fdbf8b2c9ecc58dd028fc21465 -->

What's Changed

• feat(web-shell): add session token usage panel by @ytahdn in https://github.com/QwenLM/qwen-code/pull/9988
• feat(core): accept cross-session messages behind an inbound gate by @qqqys in https://github.com/QwenLM/qwen-code/pull/9576