Sign in
← AI PulseSep 15, 2026

Deep · news · Multi-source brief

Claude Code Introduces Plugin Evaluation Tool and Desktop Pane Pop-Outs

Claude Code has released a new claude plugin eval tool for testing plugins and enabled users to pop out Desktop panes into separate windows, according to its Week 37 release notes.

By Illumora Editorial

Source · Sep 15, 2026, 12:38 AM · On Illumora · Sep 15, 2026, 12:43 AM

Video from the primary source — playable here so you can stay on Illumora.

Synthesized from multiple allowlisted primaries on the same event. Lanes →

Brief drafted by Illumora’s editorial model from the linked primary source. Ops desk reviews flagged pieces. How we write →

Read the source →Claude Code What's New + releases — Week 37 · September 7–11, 2026 - Claude Code Docs
Save

Claude Code, a product from the publisher, has introduced claude plugin eval, a new tool designed to test plugins. This feature, detailed in the Week 37 release notes (September 7–11, 2026), allows developers to run their plugins against a suite of test cases and score the results. Concurrently, the Claude Code Desktop application now permits users to pop out any pane into its own window.

Key Points

  • The claude plugin eval tool runs a plugin against a suite of test cases and scores the results.
  • By default, claude plugin eval also runs each test case without the plugin to show its contribution.
  • claude plugin eval init helps draft the test suite by asking for desired results and proposing test cases and scoring checks.
  • Every run of claude plugin eval and every check using a second model judge involves a real model call on the user's account.
  • The Claude Code Desktop app allows users to pop out any pane, such as the diff or terminal, into its own window.
  • The Week 36 release notes (August 31 – September 4, 2026) indicate that Claude Fable 5.1 is available in Claude Code with a 1M-token context window.
  • Claude Fable 5.1 is selected by the fable alias in Claude Code, while fable still selects Fable 5 in Claude apps gateway sessions.

Context

According to the Week 37 release notes, the claude plugin eval tool is initiated by running claude plugin eval init from the plugin's root directory. This command guides the user through creating a test suite. Once the suite is ready, users can exit the init session and run claude plugin eval . to score all cases. A summary table is printed in the terminal, and a detailed report is generated in report.html under evals/results/. The Week 36 release notes also mention that Claude Fable 5.1 requires v2.1.257 or later. Additionally, claude self-hosted-runner --drain-marker-file <path> has been added, allowing the runner to report its exit as a host drain when the specified file exists during a SIGTERM drain, according to the changelog.

Why It Matters

These updates provide developers with enhanced tools for plugin development and testing within the Claude Code environment. The claude plugin eval tool offers a structured method for assessing plugin performance and impact, while the Desktop pane pop-out feature improves workflow flexibility. The availability of Claude Fable 5.1 with a larger context window also expands the capabilities available to users.

What To Do

  • Test plugins using the claude plugin eval command to assess their performance and contributions.
  • Utilize claude plugin eval init to draft and refine test suites for plugins.
  • Note that each evaluation run and model-judged check consumes model calls from your account.
  • Experiment with popping out Desktop panes in the Claude Code Desktop app to optimize screen real estate and workflow.
  • Check the Claude Code documentation for specific version requirements, such as v2.1.257 for Claude Fable 5.1.