aws/agentcore-cli

What it solves

Building AI agents on Amazon Bedrock AgentCore normally means wrestling with several AWS services, setting up execution roles, and handling streamed responses. AgentCore CLI removes that friction by providing a single terminal tool that handles these details, so you can create, deploy, and invoke agents with a few commands.

How it works

The tool comes in two modes: a scriptable CLI that outputs JSON for automation (CI, scripts, coding agents) and an interactive TUI for guided workflows. You start by creating a project, which generates a spec file (agentcore.json) and a CDK app that defines your resources. You can then add agents: a Harness (managed, YAML-configured) or a Runtime (your own code, scaffolded from templates or added manually). Deployments are handled via CDK, and you can extend the generated stack to give agents access to other AWS resources like DynamoDB tables. Commands cover the full lifecycle: create, add, remove, deploy, invoke, logs, traces, evaluate, and more.

Who it's for

Developers and teams building production AI agents on AWS who want a fast, terminal-friendly workflow without deep knowledge of the underlying AWS APIs. It suits both those who prefer scriptable automation and those who like interactive prompts. Requires Node.js 20.12+ and AWS credentials.

Highlights

• Dual interface: scriptable CLI with --json output and an interactive TUI
• Supports two agent types: managed Harnesses (YAML) and custom Runtimes (your code)
• Generates a CDK app you can extend to grant agents access to other AWS services
• Includes commands for evaluate, feedback, traces, and memory management
• Templates for Python and TypeScript agents with multiple model providers (Amazon Bedrock, Anthropic, OpenAI, etc.)