What is Bedrock AgentCore?

It is an AWS managed service for rapidly deploying production-grade AI agents. It became generally available on June 18, 2026, and its two APIs—CreateHarness and InvokeHarness—let you spin up an agent with isolation, memory, observability, and scaling in minutes.

What components does it include?

It integrates seven primitives: Runtime, Memory, Gateway, Browser, Code Interpreter, Identity, and Observability. It also supports multi-model use, letting you switch between Bedrock, OpenAI, Gemini, and LiteLLM while preserving context.

Pricing is based on active usage: $0.0895 per vCPU-hour and $0.00945 per GB-hour. The travel company TUI Group was cited as an example, with over 130 people building agents simultaneously at a workshop of more than 500 participants.

Amazon Bedrock AgentCore Goes GA: From Days of Setup to a Production Agent in Minutes

Amazon Bedrock AgentCore harness became generally available (GA) on June 18, 2026, giving developers a managed service for shipping production-grade AI agents fast. The headline change is that work which once took days of infrastructure setup now happens in minutes through just two APIs—CreateHarness and InvokeHarness. It is a product aimed squarely at the realization that the bottleneck in agent development was never the idea, but the "plumbing" of infrastructure.

What Launched

Bedrock AgentCore harness is a managed service that helps developers deploy AI agents to production while minimizing infrastructure work. After a preview in April 2026, it reached general availability on June 18.

Using it is simple. The core functionality is wrapped behind two APIs—CreateHarness and InvokeHarness—so two calls are enough to spin up an agent that already has isolation, memory, observability, and scaling built in.

The Seven Core Components

AgentCore brings together seven primitives: Runtime (sandboxed execution), Memory (conversation and context retention), Gateway (tool connectivity and authentication), Browser (web interaction), Code Interpreter (running Python and Node), Identity (credential management), and Observability (CloudWatch tracing).

Combined, they handle the fundamentals of running an agent in one shot. The architecture spares you the effort of building and wiring each piece separately.

Multi-Model Support

You can switch models within a single session. Because you can move between the Bedrock, OpenAI, Gemini, and LiteLLM providers while preserving context, you can mix and match the best model for each task.

Declarative configuration is another highlight. You can set up the browser, code interpreter, MCP server, and gateway without separate adapter code, then "export" the configuration to Strands-based code—letting it evolve without structural rewrites.

What It Solves

Previously, teams burned days on infrastructure plumbing before they could even validate an agent idea. Provisioning compute, configuring storage, managing secrets, and building containers were all prerequisites.

AgentCore compresses that process from days to minutes. Two API calls launch a production-grade agent with isolation, memory, observability, and scaling baked in.

Pricing and Adoption

Billing is based on actual usage. It runs $0.0895 per vCPU-hour and $0.00945 per GB-hour, charged only for active consumption.

The scale of adoption was also showcased. At a workshop with more than 500 participants, the travel company TUI Group had over 130 people building agents simultaneously. Twilio, VTEX, and Fujisoft were also cited as examples.

What It Means for Us

This launch shows how quickly the barrier to "building an AI agent" is falling. Once infrastructure is standardized, the competitive edge shifts to "what you tell the agent to do"—the goals, the tools, and the evaluation.

The opportunity is especially real for small teams. When days of infrastructure work disappear, even a lean group can rapidly experiment with and ship production agents.

References: AWS — Amazon Bedrock AgentCore harness is now generally available (2026.6.18)