Anthropic expands Claude Platform on AWS with managed agent webhooks, multiagent orchestration, and self-hosted sandboxes

Anthropic on May 29, 2026 extended its Claude Platform on AWS with three significant capabilities: webhook support for Claude Managed Agents, multiagent orchestration, and self-hosted sandboxes. The update brings AWS deployments to feature parity with the primary Claude API for agent-based workloads.

What's new

Three capabilities—all previously available on the first-party Claude API—are now available on Claude Platform on AWS:

• Webhooks for Claude Managed Agents: Subscribe to session and vault lifecycle events via HTTP callback
• Multiagent orchestration: Run agent networks where one Claude instance coordinates others in structured workflows
• Self-hosted sandboxes: Execute tool calls in your own compute environment rather than Anthropic-managed infrastructure

Alongside these features, Anthropic added new IAM actions and the AnthropicSelfHostedEnvironmentAccess managed policy, giving security teams a standard AWS Identity and Access Management control point for managing which services and principals can interact with each capability.

Context

Claude Platform on AWS launched on May 11, 2026—just 18 days before this update—bringing the Claude API to Anthropic-managed infrastructure accessible through AWS with native AWS billing and IAM authentication. The platform gives AWS customers access to the Messages API, Files API, Message Batches API, and Claude Managed Agents without needing a separate Anthropic account.

The May 29 update fast-tracks AWS to current first-party feature parity in the agentic domain. Webhooks had been available on the main Claude API since May 6, 2026; self-hosted sandboxes and multiagent orchestration since May 19.

Self-hosted sandboxes are particularly relevant for enterprises that cannot allow tool execution to leave their infrastructure. Rather than tools running in Anthropic's cloud sandbox, they run in a customer-controlled environment. Anthropic added the new managed policy to make this configuration auditable through standard AWS IAM tooling.

Why it matters

AWS is the dominant cloud for enterprise AI workloads. By matching first-party API features rapidly on the AWS deployment path, Anthropic lowers the barrier for large enterprise teams that standardize on AWS billing, identity management, and VPC networking. Those teams can now build and run full multiagent Claude workflows without routing any traffic outside the AWS trust boundary.

Self-hosted sandboxes make Claude Managed Agents viable for regulated industries—finance, healthcare, government—where the tool execution environment is subject to compliance constraints. Organizations that need tool calls to stay within a certified environment, hit internal databases, or write to internal systems now have a compliant path to do so.

The combination of multiagent orchestration and self-hosted sandboxes specifically unlocks a class of complex automation that enterprise security teams were previously blocking: agent workflows that coordinate multiple Claude instances, call internal tools, and write to internal systems, all running in infrastructure the organization controls.