Amazon Bedrock's New Console Speaks Both OpenAI and Anthropic SDKs!

Hey everyone, it's Shii-chan! Today's news is a full makeover for the Amazon Bedrock console. It's reorganized so developers can go from picking a model all the way to production in one flow, and I'm so excited about it!

AWS Blog aws.amazon.com

What was announced?

This one comes from the AWS Blog. Amazon Bedrock introduced a new console experience built for its next-generation inference engine. The new console is optimized for an endpoint called bedrock-mantle, which supports the latest GPT, Claude, and open-weight models.

The best part is the API compatibility. It supports the OpenAI Responses API, the OpenAI Chat Completions API, and the Anthropic Messages API. So whether you come from the OpenAI side or the Anthropic side, you can call Bedrock with the SDK you already know.

The story so far

Until now, choosing a model meant hopping between docs: checking feature tables, looking up context windows, and opening quota calculators. Evaluating models, reviewing usage, and configuring the API also tended to be separate tasks.

The new console pulls all of that scattered work together under a single unit called a project.

What changes

There are three big highlights.

• New model card: you can browse the whole model catalog and compare models side by side on capabilities, modality support, context window, and applicable service quotas. You can line up as many as 3 models in a single view.
• Project-based work: you can create a project and combine running evaluations with reviewing usage insights in one workflow that mirrors how you build a generative AI application.
• Live documentation: code samples, SDK snippets, and API references are automatically prefilled with your project variables. In the words of the original post:

You can copy a snippet straight from the console into your application and run it without modification.

Dive Deep

Let's look a bit closer.

The dashboard shows inference requests and errors by recent date ranges, recently used models, and your project list. Once you create a project, assign models, and configure API keys, you can start sending inference requests in minutes.

You also get fine-grained token visibility: total token usage, token usage per minute, inference requests per minute, and tokens per inference request. That can guide your model choice and prompt optimization.

For evaluation, you can put up to 3 models side by side and compare their responses to the same prompt.

When you build an app, the Getting started flow lets you migrate existing code, start a new app with the Anthropic or OpenAI SDK, or connect an AI coding assistant to Bedrock. Pick your SDK, programming language, and authentication method, and it generates the environment variable code to test in your terminal or save to a .env file.

I love the Clients section. Choose an AI coding agent such as Claude Code, Cline, Codex, Cursor, or OpenCode, and it walks you through installing it, using your AWS IAM credentials or a Bedrock API key, setting environment variables, and routing each agent's requests through Bedrock.

For the API protocol, you can pick the Anthropic API Protocol (Messages API) for Claude, or the OpenAI API Protocol (Responses API). The references auto-fill with your project's selected model ID, Region, bedrock-mantle endpoint URL, and API key reference, and they update in place as you change settings.

By the way, the fully managed features like Agents, Knowledge Bases, Guardrails, fine-tuning, and the InvokeModel and Converse APIs still live in the existing Bedrock console on the bedrock-runtime endpoint.

Wrap-up

• Amazon Bedrock now has a new console experience built for the bedrock-mantle endpoint
• It supports the OpenAI Responses / Chat Completions APIs and the Anthropic Messages API, so you can use the SDK you already know
• The highlights are side-by-side model comparison, project-based evaluation and usage insights, and live docs that auto-fill your variables
• It even offers guided setup for AI coding agents like Claude Code, Codex, and Cursor
• Available in bedrock-mantle Regions from US East (N. Virginia, Ohio) to Tokyo, Europe, and South America

If you want to hit Bedrock with the OpenAI or Anthropic SDK, or route all your AI coding agents through Bedrock, this new experience is right up your alley!