Amazon Bedrock
A single AWS service to call AI models from Anthropic, Meta, Amazon and others.
Bedrock is the AWS service for using generative AI models through an API. Instead of running a model yourself, you send a prompt and get the answer back, paying per token. The catalog includes Claude from Anthropic, Llama from Meta, Mistral, Amazon's own Nova models and more, all behind the same API, so trying a different model is a one-line change.
What AWS adds on top of the models: your data stays in your account and your Region (it is not used to train anyone's models), access is controlled with IAM, and there are building blocks for the common patterns: Knowledge Bases for RAG (letting a model answer from your own documents), Agents that can call tools and APIs, Guardrails that filter topics and personal data, and fine-tuning for some models.
Typical example: a bank already on AWS builds an internal assistant that answers questions from its policy manuals. The documents live in Amazon S3, Bedrock indexes them into a Knowledge Base, and Claude writes the answers. Nothing leaves the company's AWS account, which is usually the reason Bedrock gets picked over calling a model provider directly.
Keep in mind
- Pay per token; no servers or GPUs to manage.
- Several model providers behind one API, including Claude.
- Your prompts and data are not used to train the models.
- Built-in pieces for RAG, agents and content guardrails.
