Claude Code v2.1.225: Gateway Spend Warnings Now Name the Cap, Agent Trust Prompt Added

Claude Code v2.1.225 makes gateway spend-limit messages actionable — you now see which cap was hit, when it resets, and the operator's note — instead of a generic "limit reached."

Two additions in v2.1.225 (Aug 7), no feature changes in v2.1.226 (Aug 8):

Gateway spend-limit messages. When a Claude Apps Gateway spend cap is reached, the usage warning now displays the cap name, its reset time, and any message the operator configured — so a CI job that stalls on a budget limit tells you exactly what hit and when it clears, rather than sending you to the gateway dashboard to find out. Requires gateway version 2.1.225+.

Workspace trust prompt for claude agents. The claude agents command now asks whether to trust an untrusted directory before running, matching the existing behavior of claude. This prevents an agent session from silently operating on code outside its intended scope when launched from an unfamiliar or shared directory.

Why it matters: If you're running Claude Code through the Claude Apps Gateway (which adds SSO and per-user spend caps for teams on AWS or Google Cloud), the improved spend message removes the main friction point when a budget cap trips in CI: you no longer need to context-switch to the gateway dashboard to understand what happened and when you can retry.

Sources: Claude Code changelog — code.claude.com · Release v2.1.225 — github.com/anthropics/claude-code · Claude Apps Gateway — byteiota.com