The Claude Code changelog lists version 2.1.323 as its newest release. Two changes matter to anyone commissioning software built with AI coding tools.
Code Mode (experimental, opt-in)
Claude writes JavaScript to call its tools, rather than calling them one at a time. It is switched off unless someone enables it, and the changelog labels it experimental. The changelog does not state performance or cost effects, so treat any such claim as unverified.
Plugin version ranges
Plugins can now declare a supported range of Claude Code versions. A plugin outside that range is not installed, updated or loaded. The same release adds TypeScript type checking for plugin hooks in claude plugin validate, reported as warnings, with a --strict option that fails on them.
Also in the release
A progress heartbeat every 30 seconds while hooks run long in non-interactive and SDK sessions, plus fixes to sign-in, memory, permission prompts, worktrees and several Bash permission checks.
Why it matters
Plugins and hooks are how teams bolt checks and company rules onto an AI coding tool. Version ranges mean a tool update is less likely to silently break those checks, and a plugin that no longer fits will be blocked, not half-working. It is a maturity signal: AI coding setups are being managed like any other piece of engineering infrastructure.
So what
Ask your supplier which extensions and rules govern their AI tooling, and who owns updating them. Our AI-assisted development service makes that part of the delivery plan. To scope a project, start here.