Image: Anthropic / claude.com Claude Code 2.1.203 Through 2.1.207: Bedrock Gets Opus 4.8 and Auto Mode Ships to Enterprise Clouds
Five releases in four days bring auto mode to AWS Bedrock, Google Vertex, and Azure Foundry, plus terminal rendering fixes, tighter security defaults, and a new /cd shortcut.
Anthropic shipped five Claude Code releases between July 7 and July 11, pushing auto mode to enterprise cloud platforms and clearing up a handful of annoying bugs along the way. Here is what changed.
Auto Mode Now Default on Bedrock, Vertex, and Foundry
The headline change in 2.1.207 (July 11) is that auto mode is now available without any opt-in on AWS Bedrock, Google Vertex AI, and Azure Foundry. Previously, teams using Claude Code through these managed cloud providers had to set an environment variable to activate auto mode. That flag is now gone. Auto mode turns on the same way it does on direct Anthropic plans.
Alongside that, Bedrock now defaults to Claude Opus 4.8, matching what direct Anthropic subscribers get for heavy agentic work.
Terminal Fixes in 2.1.207
Streaming very long content into the terminal, including big tables, lengthy code blocks, and dense prose, was causing Claude Code to freeze and making keystrokes lag. That is fixed in 2.1.207. If you have been working around this by splitting output or avoiding large responses, you can stop.
The same release fixes spurious prompt-injection warnings that were firing on ordinary system-generated messages, and patches a bug where auto-updater was overwriting custom launcher scripts.
Security Tightening in 2.1.205
Two notable security changes landed in 2.1.205 (July 8).
First, auto mode now blocks any session transcript tampering. The rule is built in and cannot be overridden through project instructions.
Second, auto mode now asks for confirmation before running rm -rf on a variable it cannot resolve from context. If Claude is about to delete something and the deletion path contains a variable it cannot evaluate, it stops and asks. This closes a class of accidental file destruction that affected a number of users in agentic workflows.
The same release also fixes a bug where messages were silently dropped when a session hit the --max-turns limit.
Quality of Life in 2.1.206 and 2.1.203
2.1.206 (July 10) added directory path suggestions to the /cd command, making it easier to navigate large repos from the chat panel. It also changed /commit-push-pr to auto-allow git push to a configured push remote without a separate permission prompt, which removes one of the most common interruptions in a common workflow.
Background agents now upgrade themselves after a Claude Code update without requiring the main session to restart.
2.1.203 (July 7) added a visible login expiry warning so sessions do not fail silently, introduced a grey pause badge in the status bar that shows the active permission mode at a glance, and passed session working directories through to the MCP roots/list response.
That last change matters for teams using MCP tools that need to scope their file access to the current project root.
What’s Still In Progress
Anthropic has not yet shipped the /checkup command mentioned in earlier changelogs, which was described as a way to audit installed skills, MCP servers, and plugins for context cost and suggest CLAUDE.md trimming. It appears this is still being tested internally.
Staying current: npm install -g @anthropic-ai/claude-code@latest or let auto-update handle it.
Sources: GitHub Releases | Releasebot | DevelopersIO