AWS Bedrock's New Console: One Interface for Anthropic and OpenAI-Compatible Models
Cloud Platform EngineeringAWS just rebuilt the Bedrock console around a reality that took longer than expected to acknowledge: teams don’t use one model. They use Claude for reasoning, GPT-4o for compatibility, and whatever specialized model fits the task — and they need a single interface to manage all of it.
What Happened
Amazon Bedrock launched a refreshed console experience built on the bedrock-mantle inference engine. The console now supports both the Anthropic Messages API and the OpenAI Chat Completions and Responses APIs from a single interface — meaning teams can use existing Claude SDK patterns or OpenAI SDK patterns against Bedrock’s model catalog without rewriting client code.
Key features in the update:
- Model Catalog: Browse and compare up to 3 models simultaneously across capability descriptions, context windows, pricing, and regional availability
- Project-Based Workflow: Organize evaluations and monitor token consumption, inference request volume, and error rates per project
- Live Documentation: Auto-populated code samples that update based on the selected model and project variables — copy-paste deployment without leaving the console
- Agent Integration: Native connection support for Claude Code, Cline, and Cursor, routing their inference requests directly through Bedrock’s engine
The console is live across 12+ AWS regions spanning US, Europe, Asia Pacific, and South America.
Why It Matters
The dual-API support is the most practically significant change. Enterprise teams that started with OpenAI-compatible code — the majority of production AI codebases in 2024–2025 — can now run that same code against Claude and other Bedrock models without client-layer rewrites. The incumbent assumption that “OpenAI-compatible” means OpenAI-hosted is being replaced by a marketplace model where the API interface and the model provider are decoupled.
For platform teams that have already invested in AWS Bedrock’s agentic architecture, the console improvements reduce the operational overhead of managing multi-model workflows. The project-based usage monitoring in particular addresses a common pain point: understanding cost and token distribution across different models and evaluation runs without building custom dashboards.
What This Means for Builders
The Claude Code and Cursor direct integration is worth evaluating for teams that want to route AI-assisted development through enterprise infrastructure rather than consumer API endpoints. Bedrock provides audit logging, IAM-based access control, and data residency guarantees that consumer endpoints don’t — meaningful for regulated industries or teams with data handling requirements.
The simultaneous model comparison (up to 3 at once) changes how model selection decisions get made. Instead of sequential A/B evaluation across separate sessions, teams can run prompts across Claude, GPT-4o-equivalent, and a specialized model in parallel and compare structured output directly.
The So What
The new Bedrock console is the clearest signal yet that AWS is positioning Bedrock as the enterprise-neutral inference layer — not the place to run Claude or GPT-4o specifically, but the managed infrastructure that can run whichever model the task needs. For platform engineers, the right question is whether Bedrock’s control plane and compliance guarantees justify routing through AWS versus calling model provider APIs directly. For most enterprise contexts, the answer is trending toward yes.
Source: AWS Blog
Content created with AI assistance and reviewed for accuracy.
Join the conversation
Stack Insiders is our free community for readers who want to go deeper — share resources, ask questions, and connect with others across every vertical we cover.
Join Stack Insiders →