Anthropic Just Made Claude on AWS a Real Platform. Full Feature Parity, IAM Auth, CloudTrail Logs.
Anthropic announced the Claude Platform on AWS yesterday. This is not the year-old Bedrock integration. The pitch is one line — every API feature Anthropic ships on the native Claude API now ships at the same time on AWS, with AWS IAM authentication, CloudTrail audit logging, and AWS unified billing including commitment retirement. Generally available across most AWS commercial regions, with global and US inference geographies.
The feature list is where this gets serious. Claude Managed Agents in beta — deploy agents at scale without standing up your own orchestration. Advisor strategy in beta. Web search and web fetch as first-class tools. Code execution for Python analysis. Files API. Skills in beta — the same skills primitive shipping under the Anthropic Skills protocol. MCP connector in beta for hooking remote MCP servers. Prompt caching, citations, batch processing. Opus 4.7, Sonnet 4.6, and Haiku 4.5 all available.
The structural piece is the simultaneous-ship commitment. Bedrock and similar managed-model platforms have historically lagged the native API by months on new capabilities — the whole reason teams kept dual-routing through Anthropic direct. The new commitment is parity from day zero. ReliableQuest, OpenRouter, and Emergent all wrote customer endorsements specifically calling out this point — canonical Anthropic API with AWS as the access layer.
Why this matters for the agent ecosystem — enterprise procurement is the constraint. Big customers have AWS contracts, AWS audit trails, AWS commitment retirement. Asking them to wire a second vendor relationship for Claude has been the friction point that capped Anthropic enterprise penetration. The Claude Platform on AWS makes Claude a first-class AWS service while keeping the full Anthropic surface area. Skills and MCP connector landing inside that envelope is the bet that the agent infra primitives compete for the same procurement chair as the underlying model.
This also ends the four-day Anthropic quiet window — last research drop was May 8 Teaching Claude Why, last news drop was the May 6 SpaceX deal. AWS GA was the Monday catch-up everyone was watching for. claude.com/blog/claude-platform-on-aws.
← Back to all articles
The feature list is where this gets serious. Claude Managed Agents in beta — deploy agents at scale without standing up your own orchestration. Advisor strategy in beta. Web search and web fetch as first-class tools. Code execution for Python analysis. Files API. Skills in beta — the same skills primitive shipping under the Anthropic Skills protocol. MCP connector in beta for hooking remote MCP servers. Prompt caching, citations, batch processing. Opus 4.7, Sonnet 4.6, and Haiku 4.5 all available.
The structural piece is the simultaneous-ship commitment. Bedrock and similar managed-model platforms have historically lagged the native API by months on new capabilities — the whole reason teams kept dual-routing through Anthropic direct. The new commitment is parity from day zero. ReliableQuest, OpenRouter, and Emergent all wrote customer endorsements specifically calling out this point — canonical Anthropic API with AWS as the access layer.
Why this matters for the agent ecosystem — enterprise procurement is the constraint. Big customers have AWS contracts, AWS audit trails, AWS commitment retirement. Asking them to wire a second vendor relationship for Claude has been the friction point that capped Anthropic enterprise penetration. The Claude Platform on AWS makes Claude a first-class AWS service while keeping the full Anthropic surface area. Skills and MCP connector landing inside that envelope is the bet that the agent infra primitives compete for the same procurement chair as the underlying model.
This also ends the four-day Anthropic quiet window — last research drop was May 8 Teaching Claude Why, last news drop was the May 6 SpaceX deal. AWS GA was the Monday catch-up everyone was watching for. claude.com/blog/claude-platform-on-aws.
Comments