A public collection of sample solutions, reference architectures, and reusable artifacts for Amazon Bedrock AgentCore.
Blueprints come in whatever form fits the pattern best: infrastructure as code, AgentCore CLI and AWS CLI scripts, agent code, and more over time. Each one is self-contained and runnable in your own AWS account, with a way to verify it works and a way to tear it down.
Amazon Bedrock AgentCore is a set of managed services for building, deploying, and operating AI agents at scale, including a serverless Runtime, Gateway for exposing tools over MCP, Memory, Identity, Observability, and built-in tools such as Web Search, Code Interpreter, and Browser. It works with any agent framework and any model. See the product page and the developer guide.
Scripted, end-to-end deployments driven by the AgentCore CLI and the AWS CLI. A good starting point if you want an agent running quickly.
| Blueprint | What it shows | AgentCore features |
|---|---|---|
| SampleAgent | Deploys a Strands agent to AgentCore Runtime or AgentCore Harness with one script, then wires in Memory and a Gateway with MCP tool targets. Includes cleanup scripts for both paths. | Runtime, Harness, Memory, Gateway, Observability |
Repeatable deployments you can review, version, and plug into a pipeline.
| Blueprint | What it shows | AgentCore features | Tooling |
|---|---|---|---|
| web-search-gateway | A Gateway exposing the AWS-managed Web Search tool over MCP with SigV4 (AWS_IAM) inbound auth and least-privilege roles. A Python MCP client proves it works without deploying an agent. |
Gateway, Web Search | Terraform |
| agentcore-private-websearch-tf | A Strands agent on Runtime inside a VPC with no internet or NAT gateway. It reaches Bedrock and the Web Search Gateway only over AWS PrivateLink, with endpoint policies scoped to the runtime role. | Runtime, Gateway, Web Search, Observability | Terraform, Docker |
More blueprints are on the way.
You'll need:
- An AWS account with access to Amazon Bedrock AgentCore in a supported Region. Some features, such as Web Search or Harness, are available in a subset of Regions.
- The AWS CLI, configured with credentials that can create the resources the blueprint provisions.
- The tooling listed in the blueprint's own README (for example the AgentCore CLI, Terraform, Docker, or Python).
Then clone the repo and open the blueprint you want:
git clone https://github.com/berry2012/agentcore-blueprints.git
cd agentcore-blueprints/<path-to-blueprint>Each blueprint's README walks through deploy, verify, and cleanup.
Contributions are welcome. See CONTRIBUTING.md for how blueprints are organized, what each one needs, and the checks that run on pull requests.
These blueprints are reference samples, not production-ready deployments. Before using them beyond a proof of concept, review the IAM policies and auth settings they create, confirm Region and service availability, and check the pricing of every service involved. Most blueprints create resources that incur charges while they exist or per invocation, so run the documented cleanup step when you are done.
This project is licensed under the MIT License. See LICENSE for details.