Claude Code 2.1.174: Token Tracking and Model Guardrails
We break down the revamped /usage command in Claude Code v2.1.174, including detailed token attribution across skills, plugins, subagents, and cache misses. Then we cover v2.1.175’s enforceAvailableModels setting for tighter enterprise model control, plus a few handy terminal and exit-time fixes.
Is this your podcast and want to remove this banner? Click here.
Chapter 1
Tracking Token Spend and Enforcing Model Guardrails in Claude Code v2.1.174 and v2.1.175
Lachlan Reed
Welcome to the show everyone! I'm Lachlan Reed, here with James Turner, and we're recording directly from my backyard shed today. A big thank you to Jellypod for making this episode possible. Now, James, mate, if you've been running the new Claude Code CLI tool, you know how quickly those API costs can run away from you. But the v2.1.174 release has completely overhauled the `/usage` command.
James Turner
Oh, absolutely! [excited] It's not just a basic prompt-and-completion counter anymore. They've actually broken down the token attribution so you can see exactly which custom skills, plugins, and background subagents are eating your budget over 24-hour and 7-day windows.
Lachlan Reed
And it even flags cache misses! [chuckles] Which is massive for anyone trying to optimize their prompt caching. If you're paying full price for a massive codebase context because of a minor edit, `/usage` will show you that exact leak.
James Turner
That cache miss tracking is a lifesaver. [matter-of-fact] But what really caught my eye in the very next dot release, v2.1.175, is this new configuration flag: `enforceAvailableModels`.
Lachlan Reed
Ah, yes, the guardrails! [laughs] Because before this, any cheeky developer in your team could just jump into their local project config and swap out the designated model for a more expensive one, or completely bypass the company's default fallbacks.
James Turner
Exactly. They'd bypass the default and suddenly you're paying for Opus when you wanted Haiku. With `enforceAvailableModels` turned on, the local system strictly clamps down on model drift. It locks the model list to what the admin defined—no widening allowed.
Lachlan Reed
It's brilliant for enterprise control. Now, they also slipped in some really nice quality-of-life fixes in this cycle. They finally solved those erratic terminal scroll issues by introducing `wheelScrollAccelerationEnabled`. I don't know about you, but my terminal was jumping around like a startled kangaroo when scrolling through long outputs.
James Turner
[laughs] Mine too! It was unusable on larger outputs. And thank goodness they fixed the exit lag. If you interrupted a shell command with a Control-C, the CLI used to hang for a painful three to four seconds before returning control. Now it's instant.
Lachlan Reed
A massive relief for developers who can't stand a sluggish terminal. That's a wrap for this quick update. We'll catch you in the next one!
James Turner
See ya!