AgentCore
Build and deploy secure, production-grade AI agents using AWS Bedrock's managed orchestration framework.
At a Glance
About AgentCore
AgentCore is a managed agent orchestration framework built into Amazon Bedrock. It enables developers to build, deploy, and scale AI agents that use foundation models and external tools to carry out complex tasks. Agents are defined with a declarative JSON schema, and AgentCore handles secure orchestration, memory, identity, observability, and multi-step reasoning across tools and APIs.
Agents can access tools like code interpreters, search, APIs, and vector databases, and support live data integration, function calling, custom routing via gateways, and per-user memory. Developers can monitor agent behavior via integrated observability, enforce identity policies with role-based access control, and deploy production-grade agents without managing infrastructure.
AgentCore is part of AWS Bedrock and supports model-agnostic orchestration, with support for Claude, Titan, and other foundation models. It's designed to support real-time LLM workflows across enterprise use cases.
Community Discussions
Be the first to start a conversation about AgentCore
Share your experience with AgentCore, ask questions, or help others learn from your insights.
Pricing
Usage-based via Bedrock
Usage-based via Bedrock plan with Billed per invocation and model usage and Pricing varies by foundation model and tools used.
- Billed per invocation and model usage
- Pricing varies by foundation model and tools used
Capabilities
Key Features
- Declarative agent schema with tool orchestration
- Built-in memory system with per-user context
- Gateway system for dynamic API routing
- Code Interpreter tool for computation and analysis
- Observability with detailed tracing and metrics
- Identity and policy enforcement using roles and scopes
- Supports both system tools and developer tools
- Live data and tool chaining for multi-step reasoning
Integrations
Demo Video
