Fetching from the wire…
01
02
03
04
05
06
07
08
Public story · 2026-07-16 · high
v2.1.211 stops Bedrock, Vertex, Mantle, and Foundry from rebilling cached system context as full-price input tokens.
Why now: The fix landed in the v2.1.211 changelog entry that entered coverage on July 16.
Claude Code shipped v2.1.211 with a fix for a prompt-caching billing regression that hit four gateway deployments: Bedrock, Vertex, Mantle, and Foundry. Trailing system context was getting rebilled as fresh input tokens instead of hitting the cache, per the changelog.
Prompt caching is the biggest cost lever in any long agent session. System prompts, tool definitions, and accumulated context are supposed to get cached and billed at a fraction of the input rate on repeat calls. The regression quietly defeated that for anyone running Claude Code through one of those four gateways. Full-price input tokens landed where cached-rate tokens should've been.
The changelog doesn't say how long the regression window was open or how much it inflated bills. Check your Bedrock, Vertex, Mantle, or Foundry invoices against your expected cache-hit ratio. Anthropic-direct usage isn't named in the fix.
The same release floors PreToolUse hook decisions so an "ask" verdict can't get silently overridden, per the changelog. It also fixes simultaneous logout across parallel sessions after wake-from-sleep. Worth a re-audit of your hook configs if you run gated tool permissions.
Each link below shares sources, entities, or timing with this story.
Go check your API spend. Right now. I'll wait. Buried in the v2.1.211 changelog is a regression fix for prompt caching on Bedrock, Vertex, Mantle, and Foundry, where trailing system context was being billed as fresh input tokens. Cached context. Re-billed at full input rate. O...
Cross-session messaging landed in v2.1.224+ (macOS and Linux only, not on Bedrock/Vertex/Foundry) and hit the HN front page at 160 points. Each session binds an inbox socket restricted to your OS user, exported to hooks as CLAUDE_CODE_MESSAGING_SOCKET and shown as Peer address...
The 2.1.233 release added an opt-in setting on Anthropic upstreams in the apps gateway that sends the signed-in user's identity as request headers (Claude Code Changelog). This closes the gap teams hit when a shared gateway makes every request look identical and per-developer...
The Write tool changed in 2.1.228 to match Edit's rules; older models still require a prior Read (Claude Code Changelog). The read-before-write guard was a cheap net against blind clobbering and it's now model-dependent rather than universal. Go re-check any hook or permission...
It now dedupes local CLAUDE.md files against checked-in ones, proposes trimming content it can derive from the codebase, ranks unused skills, MCP servers, and plugins by context cost, and flags slow hooks. Background subagents run by default and, when launched from claude agen...
Go rotate a key. I'll wait. Claude Code 2.1.246, released August 25, lists this in its changelog: a fix for "telemetry and metrics requests to Anthropic carrying the API key configured for a third-party gateway (ANTHROPIC_BASE_URL); a credential is now only sent to its own hos...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.