Qwen Code 0.26.2 adds session controls, workspace features and agent-loop experiments

Qwen Code 0.26.2 brings a substantial set of changes to the AI coding tool, including session permission rules, durable mid-turn message queuing and an experimental shared agent loop. The release also hardens several command-line paths and adds workspace capabilities, making this a practical update for people building with or operating Qwen Code.

Alibaba Qwen Watch analysis

What happened

The Qwen Code release notes list changes across managed sessions, the CLI, web shell and agent runtime. Among the additions: permission rules and automatic approval for managed sessions; durable queuing for user messages sent mid-turn; and an experimental shared agent loop that can also run headlessly. The release adds support for creating web-shell workspaces from a Git URL and uploading files to managed-agent workspaces.

Other changes include stronger protection against project environment variables affecting serve-token resolution, improved preservation and restoration of OpenTUI sessions, and fixes to memory, runtime and cost-estimation behaviour. These are changes listed in the official v0.26.2 release notes, not independent performance claims.

Why it matters

This is more than a version-number shuffle. Permission rules and approval controls affect how managed agents are allowed to act; durable message queuing addresses what happens when a person sends instructions while an agent is already working. Workspace creation and uploads broaden the ways developers can provide projects and files to the tool.

The release also points to Qwen Code being developed as a broader agent platform, with experimental shared-loop support alongside its CLI and managed workflows. The notes establish that these features were added, but do not by themselves show how they perform in everyday use.

Our read

There is enough here to merit a proper look, particularly for developers using Qwen Code’s managed sessions or web shell. The most consequential changes are about control and continuity: defining permissions, handling approvals and keeping work moving when new instructions arrive. The experimental loop is intriguing, but “experimental” is doing useful work in that description. Check the release notes for the specific features relevant to your setup before upgrading a production workflow.

What to watch

• How permission rules and automatic approval behave in managed sessions.
• Whether the shared agent loop moves beyond its experimental stage.
• How durable mid-turn queues handle competing or changing instructions.
• Whether the new workspace features make web-shell workflows easier to adopt.

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

not affiliated with or endorsed by Qwen