Claude Code v2.1.237: Prompt Caching Fix for LLM Gateways + Built-in Concise Output Style

Claude Code v2.1.237 is out, with two changes that affect day-to-day usage: a fix for prompt caching when you're running through an LLM gateway or custom base URL, and a new built-in 'Concise' output style.

Prompt caching fix for gateways

If you use Claude Code with an LLM gateway (like LiteLLM or a corporate proxy) or a custom base URL, prompt caching was apparently broken in previous versions. That defeats the purpose — caching is what keeps large context windows fast and cost-effective. v2.1.237 addresses that, so sessions via these setups should see the usual cache hits again.

New 'Concise' output style

There's also a new output style called Concise. From the changelog: "Claude leads with results and skips preamble and narration, while doing the work just as thoroughly." In practice that means less filler like "I'll now analyze your codebase" and more straight-to-the-point responses. You can switch to it by running /config and selecting Concise under Output style.

This is a global setting, so once you pick it, it applies across your sessions. If you've been using custom prompts to force brevity, this is a first-class option now.

Who this is for

• Teams routing Claude Code through an LLM gateway or custom base URL — the caching fix matters for performance and token costs.
• Developers who want less narration and faster answers — the Concise style is worth trying.

The release is tagged v2.1.237 on GitHub. Update with npm update -g @anthropic-ai/claude-code or your usual installer.