Anthropic News: Claude Managed Agents Now Available on AWS with Three Key Features

[1] Claude Managed Agents: Three Features Now Officially Supported on AWS

The Platform Release Notes dated May 29, 2026 formally announced that three Claude Managed Agents features — webhooks, multiagent orchestration, and self-hosted sandboxes — are now available on Claude Platform on AWS. Along with this, a new list of IAM actions and the AnthropicSelfHostedEnvironmentAccess managed policy were published, enabling complete management of autonomous agent execution environments using native AWS permission controls. Claude Platform on AWS, announced on May 11, 2026, is a fully managed Claude API platform that supports AWS billing and IAM authentication.

Impact for engineering readers: ★★★. This makes it practical to reuse existing AWS VPCs, security groups, and IAM roles as the foundation for agent infrastructure — a directly relevant architecture for teams that need to pass enterprise security reviews. The AWS support for self-hosted sandboxes fulfills the requirements of large enterprise users who want to run code execution entirely within their own AWS accounts rather than depending on Anthropic's infrastructure. For details, see the Platform Release Notes entry for May 29, 2026 and the Claude Platform on AWS IAM Actions reference.

[2] Claude Code v2.1.156 Fixes Thinking Blocks Error with Opus 4.8

At 1:42 AM Japan time on May 29, 2026, Claude Code v2.1.156 was released. This is a targeted single bug fix addressing an API error that occurred when thinking blocks were modified while using Opus 4.8. Thinking blocks are a core component powering Opus 4.8's adaptive thinking, and the API returns 4xx errors when they are sent in a modified state from an external source.

Impact for engineering readers: ★★. An immediate update from the previous version v2.1.154 is recommended, and the release has drawn significant community attention with 42 reactions. Teams working with thinking blocks in long-running tasks or multi-turn execution environments like Dynamic Workflows should promptly update with npm install -g @anthropic-ai/claude-code. Source: claude-code v2.1.156 release notes.

[3] TypeScript SDK v0.100.1 Fixes Missing Encrypted Content During Streaming

At 12:08 AM Japan time on May 29, 2026, three packages were released simultaneously: the core package sdk v0.100.1, foundry-sdk v0.2.4, and aws-sdk v0.3.1 (all packages within the anthropic-sdk-typescript repository). The core change, introduced in PR #1025, is a fix for a bug where encrypted_content in beta compaction blocks was not preserved during streaming. Beta compaction, released with Opus 4.6 on February 5, 2026, automatically compresses long-session context on the server side — and the issue of encrypted block content being lost when received via streaming has now been resolved.

Impact for engineering readers: ★★. Products using the TypeScript SDK that combine streaming with compaction are advised to update to v0.100.1. The foundry-sdk and aws-sdk updates contain only lockfile dependency updates with no functional changes. For details, see the anthropic-sdk-typescript releases page.

[4] Python SDK v0.105.2 / v0.105.1 Adds Support for PyPI Trusted Publishing

In the early hours of May 29, 2026 (Japan time), Python SDK v0.105.1 and v0.105.2 were released in quick succession. Trusted Publishing is a mechanism that allows packages to be securely published to PyPI using short-lived credentials generated at build time, rather than storing long-lived API tokens or passwords in CI environments. v0.105.1 added the id-token permission required for Trusted Publishing to the continuous integration pipeline — a supply chain security improvement — while v0.105.2 is a corrective release commit for that release. This shifts the distribution flow away from reliance on long-lived credentials.

Impact for engineering readers: ★. Neither release changes the SDK's API specifications or behavior, so no modifications are required in code that calls the Python SDK. Version pins for products depending on the Python SDK can be updated at your discretion. Source: anthropic-sdk-python releases page.

What to Watch Next

Following the major news of Claude Opus 4.8 and the Series H funding round on May 28, the 29th was entirely focused on stabilization patches and AWS integration enhancements. In particular, the addition of three Managed Agents features to Claude Platform on AWS is worth watching as a move that fills the last mile for enterprise adoption. Going forward, key things to watch include whether AWS IAM-based permission templates will be developed as official references, when expansion to Azure or Google Cloud Platform might be announced, and how quickly the post-release patches for Opus 4.8 converge. The Clauder Navi editorial team will continue bringing you the latest official Anthropic updates.

Sources

• Platform Release Notes (2026-05-29 entry): https://platform.claude.com/docs/en/release-notes/overview
• Claude Managed Agents — Webhooks: https://platform.claude.com/docs/en/managed-agents/webhooks
• Claude Managed Agents — Multi-Agent: https://platform.claude.com/docs/en/managed-agents/multi-agent
• Claude Managed Agents — Self-Hosted Sandboxes: https://platform.claude.com/docs/en/managed-agents/self-hosted-sandboxes
• Claude Platform on AWS IAM Actions: https://platform.claude.com/docs/en/api/claude-platform-on-aws-iam-actions
• Claude Code v2.1.156: https://github.com/anthropics/claude-code/releases/tag/v2.1.156
• Anthropic TypeScript SDK Releases: https://github.com/anthropics/anthropic-sdk-typescript/releases
• Anthropic Python SDK Releases: https://github.com/anthropics/anthropic-sdk-python/releases