AWS Bedrock Agents
AWS Bedrock Agents, renamed Amazon Bedrock Agents Classic in 2026, are serverless AI orchestration services that let developers build autonomous agents that execute multi-step tasks by dynamically invoking APIs, querying knowledge bases, and reasoning through problems using foundation models. The service entered maintenance mode on July 30, 2026: existing agents continue to run and receive bug fixes and security patches, but it is closed to new customers, no new features are planned, and its model catalog is frozen. New agent projects on AWS go through Amazon Bedrock AgentCore. These managed agents leverage foundation models available in Amazon Bedrock from Anthropic, Amazon, Meta, Mistral, and other providers to process natural language instructions, break down complex requests into actionable steps, and coordinate multiple tools and data sources. AWS Bedrock Agents incorporate action groups that define available functions, knowledge bases for retrieval-augmented generation, and guardrails that ensure safe and compliant agent behavior. The service provides automatic prompt engineering, session management, and conversation memory that enable agents to maintain context across interactions while adapting to changing requirements. Advanced capabilities include function calling, API integration, and custom business logic execution that allow agents to interact with external systems, databases, and enterprise applications. AWS Bedrock Agents utilize fully managed infrastructure that handles scaling, security, and reliability while providing detailed monitoring and logging for agent performance optimization. These agents are essential for automating business processes, creating intelligent assistants, and building complex AI workflows that require reasoning, planning, and multi-step task execution within enterprise cloud environments.
Related terms
Related services: Single-agent system development, Agentic AI consulting.
Ready to put agentic AI to work?
Book a free 45-minute consultation. We'll map one real process worth automating with production-grade AI.