Amazon Bedrock AgentCore gets managed harness: a working agent in just three API calls
Why it matters
Amazon announced a managed agent harness for Bedrock AgentCore that enables deploying a fully working agent in just three API calls, without writing any orchestration infrastructure. The harness is accompanied by the AgentCore CLI covering the full development cycle and pre-built skills for coding assistants, available in preview across four AWS regions.
Amazon Web Services announced a set of new features for Amazon Bedrock AgentCore, its platform for developing AI agents. The centrepiece announcement is the managed agent harness that dramatically simplifies the path from idea to working agent.
What does the managed harness bring?
The managed harness is a managed agent runtime. Developers no longer need to write their own orchestration infrastructure — the logic for calling tools, managing conversation state, retry loops, and model integration. AWS takes care of all that.
According to Amazon’s documentation, a fully working agent can be deployed in just three API calls. That number is significant because it shows the goal is to reduce initial complexity to a minimum, similar to how Lambda functions once simplified serverless backends.
The managed harness is currently in preview and available in four AWS regions. Amazon has not announced a general availability date, meaning production use should be planned with that caveat in mind.
What do the AgentCore CLI and pre-built skills bring?
The harness is accompanied by the AgentCore CLI, a command-line tool covering the entire agent development lifecycle — from project initialization and local testing to deployment and execution monitoring. The CLI is available in all commercial AWS regions, unlike the harness itself.
The second new feature is pre-built skills for coding assistants — ready-made skills that developers can plug into their agents. This means a team building an internal coding assistant doesn’t have to write functions for reading repositories, running tests, or analyzing code from scratch.
Both the CLI and pre-built skills are available immediately in all commercial regions.
How does this differ from the earlier AgentCore announcement?
An important note for tracking: this is a new announcement from April 22, 2026 and relates exclusively to the managed harness, CLI, and coding skills. It should not be confused with the earlier announcement from April 21, when Amazon introduced AgentCore integration with the Nova Sonic voice model for omnichannel ordering.
Both announcements are part of the broader AWS strategy to position Bedrock AgentCore as a competitor to OpenAI Assistants API, Google Vertex AI Agent Builder, and Microsoft’s Azure AI Foundry. The focus on deployment speed and managed infrastructure is a clear signal that AWS is targeting developers who want agents without managing orchestration.
Sources
Related news
Anthropic: Memory for Managed Agents in public beta — AI agents that remember context between sessions
GitHub: Cloud agent sessions now available directly from issues and project views
ArXiv SWE-chat — a dataset of real developer interactions with AI coding agents in production