Image: Anthropic Claude Code v2.1.158 Brings Auto Mode to AWS Bedrock, Google Vertex, and Azure Foundry
The latest Claude Code release extends autonomous agent mode to developers running Claude through cloud provider inference platforms. You opt in with a single environment variable.
Claude Code v2.1.158 shipped with one notable change for enterprise teams: auto mode now works on AWS Bedrock, Google Vertex AI, and Azure AI Foundry, not just through Anthropic’s direct API. It supports both Opus 4.7 and Opus 4.8.
To enable it, set the environment variable:
CLAUDE_CODE_ENABLE_AUTO_MODE=1That’s the full opt-in. No additional configuration required if your Bedrock, Vertex, or Foundry credentials are already set up.
Why this matters for enterprise users
Auto mode is how Claude Code handles long-horizon tasks without interrupting you every few steps. The agent maintains context, makes decisions about what tools to use, and keeps working until the task is done or it genuinely needs input. Without it, you’re managing the agent’s decisions yourself more often.
Until this release, auto mode was only available when routing through Anthropic’s own API. A significant share of enterprise Claude Code deployments go through cloud providers instead, whether for data residency requirements, consolidated billing, or because the team is already running other workloads on Bedrock or Vertex. Those users were getting a more manual experience than direct API users.
The opt-in design makes sense here. Auto mode gives the agent substantially more latitude, and enterprise environments often have tighter controls around what automated processes can do. Setting an env var means it’s an explicit choice, easy to track in deployment configs and CI environments, and easy to audit.
What else shipped recently
The last few Claude Code releases have been moving fast. Version 2.1.157, which shipped immediately before this one, added automatic loading of plugins from .claude/skills directories. Previously, skills required marketplace installation; now any skill file dropped into that directory loads automatically.
Version 2.1.154, released May 28, was the larger milestone: Opus 4.8 as the new default model, dynamic workflows for distributing work across multiple background agents, and a fast mode price reduction on Opus 4.8 (2x standard rate for 2.5x speed). Version 2.1.156 followed to fix a thinking-block bug in Opus 4.8 that was causing API errors on certain requests.
The pattern in these releases is consistent: each version is small, ships fast, and addresses something specific. The Bedrock/Vertex/Foundry auto mode change is a single-line feature for users who’ve been waiting for it, and ships as version 2.1.158 rather than waiting for a bigger bundle.
Sources: Claude Code CHANGELOG, Releasebot Anthropic Claude Code May 2026