Source: Amazon Bedrock AgentCore — Overview

AWS's modular, framework-agnostic agent platform: Runtime, Harness, Memory, Gateway, Identity, tools, Observability, Evaluations, Policy and more.

Contents

Summary

Lists the AgentCore services, which you can use together or separately with any framework and any model, inside or outside Bedrock.

Key claims

  • Harness: a managed agent loop; give it a model, system prompt and tools in one API call; each session runs in an isolated microVM with filesystem and shell.
  • Runtime: serverless hosting for agents and tools (CrewAI, LangGraph, LlamaIndex, Google ADK, OpenAI Agents SDK, Strands), any model, MCP and A2A protocols.
  • Memory: short-term and long-term memory, shareable across agents.
  • Gateway: turns APIs, Lambda functions and services into MCP tools, and fronts existing MCP servers.
  • Identity: agent identity and auth that works with Cognito, Okta, Entra ID and Auth0.
  • Code Interpreter and Browser: managed sandboxes (Python/JS/TS; Playwright/BrowserUse).
  • Observability: OpenTelemetry-compatible tracing. Evaluations and Optimization (AI recommendations + A/B tests via Gateway).
  • Policy: deterministic rules (natural language or the Cedar-compatible Dogwood language) enforced on every Gateway tool call.
  • Registry: catalogue of agents, MCP servers, tools and skills. Payments: x402/MPP micro-payments for agents.
  • Pricing is consumption-based with no minimums.

Feeds

Amazon Bedrock

Suggest changes →