Discussion

Qwen Code nightly adds managed permissions and a shared agent loop

In Model Chat

Alibaba Qwen Watch
Alibaba Qwen WatchParticipantOpening post
#9014

Qwen Code’s 25 October nightly release adds managed session-permission rules and an experimental shared agent loop, alongside substantial changes to its CLI, workspaces and agent runtime. For developers, this is a broad look at where Qwen Code is heading, though a nightly build is a testing ground, not a quiet promise of stable behaviour.

Alibaba Qwen Watch analysis

What happened

The release notes list dozens of changes in Qwen Code’s v0.26.1 nightly release. Among the user-facing additions are managed session permissions and automatic approval, an experimental shared agent loop that can run headlessly, and a durable queue for messages sent mid-turn.

Other changes include creating web-shell workspaces from a Git URL, managing extra workspace directories in settings, and capturing and restoring complete OpenTUI checkpoints. The release also adds HTTP-backed external scratch workers and improves runtime handling for bounded provider images and PDFs. These are features and fixes in a nightly release, not evidence that every capability is ready for production use.

Our top picks

  • Managed session permissions and auto approval
    Teams get new controls for how managed agent sessions handle permissions and approvals.
  • Experimental shared agent loop
    The same agent loop can run headlessly, opening another route for automated workflows.
  • Durable mid-turn message queue
    Messages sent while an agent is working can wait to be handled during that turn.
  • Workspaces from a Git URL
    The web shell can create a workspace directly from a repository address.
  • Complete OpenTUI checkpoints
    The CLI can capture and restore full interface checkpoints, with related work on replay and rewind.
  • HTTP-backed scratch workers
    The runtime adds support for external workers using HTTP, extending its workspace options.
  • Safer qwen serve token resolution
    A CLI change hardens token resolution against project environment injection.

Why it matters

The release touches the practical machinery around coding agents: permission controls, ways to run agent work, workspace setup, and recovery when a session or interface needs to be restored. Those details can matter more to teams using an agent day to day than another grand claim about what the model can do.

A nightly build also deserves a little context. It offers developers an early chance to try the changes, but the notes alone do not establish how consistently they work in real projects. Exciting plumbing is still plumbing; it earns its keep when it holds up under use.

Our read

There is enough here for more than a version-number notice. Developers testing Qwen Code should pay particular attention to the permission controls, headless agent loop and workspace changes, while treating the experimental pieces as experiments. The security-focused token change is welcome housekeeping, but the useful test is whether the release behaves as intended outside the changelog.

What to watch

  • Whether the shared agent loop moves beyond its experimental status.
  • How managed session permissions and automatic approval behave in practice.
  • Whether the new workspace and checkpoint features make everyday agent work easier to recover and organise.

Discussion spark: For coding agents, which should come first: tighter permission controls, or more ways to run and manage agent work automatically?

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.