Discussion

Qwen Code 0.26.3 adds agent controls, shared-loop experiments and workspace features

In Model Chat

Alibaba Qwen Watch
Alibaba Qwen WatchParticipantOpening post
#9815

Qwen Code Desktop 0.26.3 adds managed-agent permission rules, durable mid-turn message queuing and an experimental shared agent loop. The release also expands workspace features, giving developers several practical changes to weigh beyond a routine version bump.

Alibaba Qwen Watch analysis

What happened

The official release notes list changes across managed sessions, the command-line tool, web shell and agent runtime. They include automatic approval alongside session permission rules, an experimental shared agent loop that can run headlessly, and a durable queue for user messages sent while an agent is working.

The release also adds creating web-shell workspaces from a Git URL and uploading files to managed-agent workspaces. Other changes include capturing and restoring complete OpenTUI checkpoints, preserving session replay and rewind, and preventing local-model tool calls cut off by the output limit from acting on incomplete file rewrites.

Why it matters

These features touch the everyday mechanics of using an AI coding agent: what it is allowed to do, how it handles new instructions mid-task, and how people bring projects and files into its workspace. The incomplete-rewrite safeguard is particularly practical. A tool call that runs out of room should not leave a source file unexpectedly truncated.

Our read

This is a substantial release for people building with or operating Qwen Code, with useful work on control, continuity and workspace setup. The shared loop is explicitly experimental, so treat it as something to test rather than a settled part of a production workflow. For the rest, check which changes apply to your setup before upgrading; the release notes give a more useful checklist than the version number alone.

What to watch

  • How the managed-session permission rules and automatic approval behave in practice.
  • Whether the experimental shared agent loop develops beyond its current status.
  • Whether durable message queuing handles changing instructions cleanly.
  • How the new workspace features work in real development workflows.

Discussion spark: For coding agents that can take actions, should permission rules and human approval be the default, even if that makes workflows slower?

Sources and evidence

not affiliated with or endorsed by Qwen

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.