Can a user override the maxEffortLevel cap in Claude Code?

No — maxEffortLevel is a ceiling, not a default. Claude Code 2.1.267 applies it on every provider including Bedrock, Vertex and Foundry, and users can still select any level below it. Set it at the top level of your settings to cover everything, or per model under modelSettings when only certain models need the limit.

When would I use --system-prompt-snapshot off?

Claude Code normally reuses the system prompt recorded with a conversation, which keeps the prompt cache warm. The --system-prompt-snapshot off flag in 2.1.267 renders the prompt fresh on every request instead, which is what you want while iterating on prompt text and need each change to take effect immediately rather than at the next session.

Does Claude Code 2.1.267 fix anything for resumed sessions?

Several things. Resuming a transcript over 5 MB no longer drops parallel tool calls and their hook output. Resuming after /compact or another slash command via -p --resume no longer inserts a spurious "Continue from where you left off." turn. Resumed sessions also stop re-rendering tool descriptions and rewriting earlier MCP tool announcements, both of which had been discarding extended thinking.

What changed in the VS Code extension in this release?

Claude Code 2.1.267 fixes the extension host hanging at 100% CPU when forking or rewinding a conversation whose saved transcript contains a cyclic parent link. Chat diff blocks now follow the active VS Code colour theme including high contrast, @-mentions stop dropping files whose paths contain spaces, and accepting an edit on a file with Windows CRLF line endings no longer fails with "String not found in file".

Anthropic · 2026-09-09 · major

Claude Code 2.1.267 — one setting caps effort on every provider

Claude Code 2.1.267 adds maxEffortLevel, a setting that caps the effort level on every provider including Bedrock, Vertex and Foundry. Users can still choose a lower level. A long run of fixes repairs prompt-cache reuse.

Claude Code v2.1.267 release page on GitHub — GitHub

Claude Code 2.1.267 lets an admin cap effort level everywhere at once and spends most of its changelog repairing prompt-cache reuse.

Quick facts

Maker	Anthropic
Version	2.1.267
Released	September 9, 2026
New setting	maxEffortLevel (top level or per model)
New flag	--system-prompt-snapshot off
Security fix	Marketplace path containment bypass on macOS and Linux

What is it?

The new maxEffortLevel setting in Claude Code 2.1.267 puts a ceiling on effort level across every provider — Bedrock, Vertex and Foundry included — and can be set at the top level or per model under modelSettings. Users keep the freedom to pick a lower level; they just cannot go above the cap. A second addition, --system-prompt-snapshot off, renders the system prompt fresh on every request instead of replaying the one recorded with the conversation.

How does it work?

Most of this build hunts prompt-cache misses. Switching models with /model used to re-send every tool definition; commit and PR attribution text now arrives as a conversation note that updates on model change instead. Sessions started with --system-prompt or --append-system-prompt record their prompt and tools once rather than re-rendering them, and mid-session MCP or plugin tools now arrive as deferred definitions on supported models instead of rewriting the tool list.

Why does it matter?

Prompt-cache breaks cost money and drop extended thinking, and this release closes the common ones: a reconnecting MCP server, a resumed session over 5 MB, a print-mode conversation continued interactively. A security fix is folded in too — a marketplace entry path containing a backslash could bypass the containment check for fetched marketplaces on macOS and Linux. Anyone running Claude Code through a gateway should also note that forwarded-identity upstreams now return a 429 as-is instead of failing over.

Who is it for?

Claude Code users and the admins who set organization policy

Frequently asked questions

Can a user override the maxEffortLevel cap in Claude Code?: No — maxEffortLevel is a ceiling, not a default. Claude Code 2.1.267 applies it on every provider including Bedrock, Vertex and Foundry, and users can still select any level below it. Set it at the top level of your settings to cover everything, or per model under modelSettings when only certain models need the limit.
When would I use --system-prompt-snapshot off?: Claude Code normally reuses the system prompt recorded with a conversation, which keeps the prompt cache warm. The --system-prompt-snapshot off flag in 2.1.267 renders the prompt fresh on every request instead, which is what you want while iterating on prompt text and need each change to take effect immediately rather than at the next session.
Does Claude Code 2.1.267 fix anything for resumed sessions?: Several things. Resuming a transcript over 5 MB no longer drops parallel tool calls and their hook output. Resuming after /compact or another slash command via -p --resume no longer inserts a spurious "Continue from where you left off." turn. Resumed sessions also stop re-rendering tool descriptions and rewriting earlier MCP tool announcements, both of which had been discarding extended thinking.
What changed in the VS Code extension in this release?: Claude Code 2.1.267 fixes the extension host hanging at 100% CPU when forking or rewinding a conversation whose saved transcript contains a cyclic parent link. Chat diff blocks now follow the active VS Code colour theme including high contrast, @-mentions stop dropping files whose paths contain spaces, and accepting an edit on a file with Windows CRLF line endings no longer fails with "String not found in file".

Try it

npm i -g @anthropic-ai/claude-code@2.1.267