Discussion

Qwen Code 0.26.6 adds agent activity views and richer usage metrics

In Model Chat

Alibaba Qwen Watch
Alibaba Qwen WatchParticipantOpening post
#10724

Qwen Code 0.26.6 adds a conversation activity view, new per-request speed and failure charts, and a broader set of agent and workflow changes. The stable release gives developers several concrete improvements to inspect, including clearer visibility into what their AI coding sessions are doing.

Alibaba Qwen Watch analysis

What happened

The Qwen Code release notes list changes across the web shell, agent runtime, desktop app and core tools. The web shell can now show conversation activity, request success rates, distinct model colours, and per-request speed and failures in token usage. The release also adds “Add to Split View” to session menus and makes GitHub channel configuration usable.

Other changes include support for running prompt-submit hooks in headless runs, a shared loop for re-entered and other turns, and fixes covering MCP tool-name collisions, LSP queries when no server is ready, and shell permission checks for 2>&1 redirection. The notes list the changes in the Qwen Code 0.26.6 release.

Our top picks

  • See conversation activity
    A new web-shell view gives developers another way to follow what is happening in a session.
  • Inspect request success and speed
    Token usage now includes request success rates, distinct model colours, speed and failures.
  • Add sessions to Split View
    A session menu option makes it easier to put a conversation into a split view.
  • Run prompt-submit hooks headlessly
    Mods can now run these hooks in headless runs, extending their use beyond interactive sessions.
  • Configure GitHub channels in the web shell
    The release describes the configuration as usable, a practical step for people working with that integration.
  • Keep conflicting MCP tools available
    A fix preserves uniquely named tools when tool names collide.

Why it matters

For people building with an AI coding agent, visibility is part of usability. Request-level speed and failure information can help explain how a session is behaving, while the activity view gives developers another way to follow its progress. The release also combines those user-facing additions with changes to agent execution and integrations.

The release notes document what changed, but they do not establish how these features perform in everyday projects. Even so, this is a substantial enough set of changes to merit a proper look, rather than a passing nod to a new version number.

Our read

The stand-out is the extra visibility into sessions and requests: useful feedback for anyone trying to work out whether an agent is making progress or merely keeping the interface busy. Developers using Qwen Code should check the notes for changes relevant to their setup. The useful test is whether the new views help diagnose real work, not whether the charts look suitably industrious.

What to watch

  • How the conversation activity view works in day-to-day sessions.
  • Whether request speed and failure metrics help developers spot problems sooner.
  • How the shared loop and headless hook changes affect agent workflows.

Discussion spark: When an AI coding agent struggles, which would help you more: a clear view of its activity, or better request-level speed and failure metrics?

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.