AWS Unveils AgentCore: A Modular Platform for Building Scalable AI Agents

• Launch Details and Vision

  AWS introduced Amazon Bedrock AgentCore at its New York Summit 2025, branding it a “tectonic shift” in software development and enterprise workflows. According to Swami Sivasubramanian (VP, Agentic AI at AWS), AI agents will redefine how software is deployed, used, and interacted with—moving from reactive tools to autonomous systems that can act on behalf of humans.

• Architecture & Modular Components

  AgentCore is a preview-stage, composable service stack comprised of:

  • AgentCore Runtime: Secure, serverless execution for long-running sessions (up to 8 hours) and asynchronous tasks. Supports full session isolation.
  • AgentCore Memory: Provides short- and long-term memory capabilities, enabling contextual learning and persistent knowledge across interactions.
  • AgentCore Identity: OAuth-based system with integrations for Amazon Cognito, Microsoft Entra ID, and Okta, allowing agents to operate securely across tools.
  • AgentCore Gateway: Converts internal APIs, Lambda functions, and external services into agent-compatible tools using Model Context Protocol (MCP).
  • AgentCore Code Interpreter: Customizable sandbox for agent-executed code, supporting data analysis, visualizations, and automated scripting.
  • AgentCore Browser Tool: Cloud-based, model-agnostic browser interface for agents to navigate, extract, and interact with websites.
  • AgentCore Observability: Tracks, logs, and monitors agent actions using Amazon CloudWatch, LangSmith, OpenTelemetry, and Datadog integration.

• Open Ecosystem Compatibility

  AgentCore supports broad framework and model interoperability:

  • LangChain, CrewAI, LangGraph, LlamaIndex, and Strands Agents SDK
  • Model-agnostic architecture (not restricted to Bedrock)
  • Meta’s Llama LLM supported via official partnership and startup program (30 companies supported with up to $200,000 in AWS credits and expert guidance)

• Pricing Model

  Free until September 16, 2025, after which consumption-based pricing applies:

  • Runtime, Browser, Code Interpreter: $0.0895 per vCPU-hour; $0.00945 per GB-hour
  • Gateway: $0.005 per 1,000 API calls; $0.025 per 1,000 search queries
  • Memory: $0.25–$0.75 per 1,000 events
  • Identity: $0.010 per 1,000 token/API key requests
  • Observability: Per CloudWatch usage

• Early Adoption Across Sectors

  Companies already deploying or integrating AgentCore include:

  • Box: High-security AI-powered content workflows using AgentCore + Strands
  • Itaú Unibanco: Personalized digital banking experiences
  • Innovaccer: Built HMCP (Healthcare Model Context Protocol) for agent interoperability
  • Epsilon: Reduced marketing campaign build time by 30%
  • Boomi: Integrating agent-based automation across enterprise environments
  • AllCloud, Accenture, Brave, CircleCI: Listed early solutions in the AWS Agent Marketplace

• Expanded AI Ecosystem

  • Kiro IDE: Released July 14. A multi-cloud, agent-specific development interface that improves prompt engineering and configuration.
  • Amazon S3 Vectors: Adds vector database capabilities to Amazon’s storage system for retrieval-augmented generation (RAG) workflows.
  • Nova Foundation Models + SageMaker HyperPod: Enhancements include serverless inference, observability for cluster diagnostics, and scalable model training.
  • TwelveLabs Integration: Semantic video intelligence now part of the Bedrock ecosystem.

• AI Agents and Tools Marketplace – Over 900 Listings

  • Launched as a dedicated category in the AWS Marketplace
  • Bundles prebuilt agents, tools, knowledge bases, governance guardrails, and implementation services
  • Early listings include providers such as Anthropic, Salesforce, IBM, PwC, Stripe, C3.ai, Perplexity, and Automation Anywhere
  • Semantic AI-powered search helps users find agents by use case (e.g., procurement, supply chain, compliance, document intelligence)
  • Supports free trials, pay-as-you-go, subscriptions, and custom enterprise pricing
  • Vendors can create private offers, leverage standardized contracts, and manage deals within AWS’s procurement flow
  • Over 99% of AWS’s top 1,000 enterprise customers already use the Marketplace

• Developer-Centric Strategy

  AWS prioritizes CLI-first and SDK-driven workflows over low-code solutions. However:

  • Feedback suggests AWS must package ready-made agentic solutions to support non-developer stakeholders.
  • The AI Marketplace addresses this by offering accessible agent bundles and prebuilt integrations for common enterprise needs.

• Observability & Accuracy

  • AgentCore Observability tracks every agent action in production environments.
  • John Balsavage (A&I Solutions) highlighted its value in driving agent reliability to 100%, beyond industry norms.
  • Addresses concerns cited by CFOs in PYMNTS Intelligence July 2025 report, which noted fear of “off-script” agent behavior causing compliance violations and payment errors

• Partnerships and Enablement

  • $100M investment in AWS Generative AI Innovation Center
  • 6,600+ AWS Academy students now receive free training access and exam vouchers
  • Joint AWS-Meta startup initiative provides credits and mentorship

• European Sovereign Cloud Expansion

  • Set to launch by end of 2025 in Brandenburg, Germany
  • Independently governed within the EU
  • Designed for public sector, regulated industries, and compliance-sensitive enterprises

• Standards and Interoperability

  • Supports Model Context Protocol (MCP) and Google’s Agent-2-Agent (A2A)
  • Facilitates secure cross-agent communication, discovery, and tool compatibility

• Competitive Landscape

  AWS positions AgentCore against:

  • OpenAI Agents SDK
  • Google Gemini + DeepMind tools
  • Anthropic’s Claude ecosystem
  • Writer’s AI HQ, Cognition’s Devin
  • Differentiators include agent modularity, observability depth, and enterprise procurement integration

• Global Availability

  Available in preview in:

  • US East (N. Virginia)
  • US West (Oregon)
  • Asia Pacific (Sydney)
  • Europe (Frankfurt)

• Developer Resources

  AWS supports adoption via:

  • GitHub repos and CLI samples
  • Official documentation
  • Discord developer community
  • SDKs and integrations with Bedrock

Why This Matters

AWS is accelerating the adoption of enterprise AI agents by building a complete and composable ecosystem—AgentCore, Kiro IDE, and the AI Agents Marketplace—combined with enterprise-friendly tools for observability, compliance, and procurement. With over 900 pre-integrated agent listings and growing vendor participation, AWS is becoming the default environment for companies transitioning from experimentation to production-grade agentic systems. As market concerns about autonomy, compliance, and cost persist, AWS’s emphasis on control, modularity, and governance offers a differentiated, end-to-end path for enterprise-grade deployment.