← all repositories
awslabs/fullstack-solution-template-for-agentcore

AWS's "vibe coding" starter kit for Bedrock agents

A forkable template that tries to let AI coding assistants handle the fullstack scaffolding so developers can focus on prompts and agent logic instead of Cognito wiring.

510 stars Python AgentsApp Builders
fullstack-solution-template-for-agentcore
Velocity · 7d
+3.6
★ / day
Trend
steady
star history

What it does

FAST is a deployable starter repo for building web apps on Amazon Bedrock AgentCore. It bundles a React frontend, Cognito auth, CDK (or Terraform) infrastructure, and a minimal multi-turn chat agent backend with tools. The pitch: fork it, run cdk deploy, then let your AI coding assistant customize everything while you stay in “vibe coding” mode.

The interesting bit

The project deliberately keeps its baseline almost insultingly simple — a word-counting text tool and a Python code interpreter — because the real product is the documentation. AWS codified security and architecture best practices into markdown files meant to be fed directly into coding assistants as context. It’s a bet that the undifferentiated heavy lifting isn’t the code, but the knowledge of how to wire Cognito JWTs through to Cedar policy enforcement correctly.

Key highlights

  • Dual infrastructure paths: CDK and Terraform are both included; the docs nudge you to delete the one you don’t use
  • Auth flow is fully worked out: Cognito → frontend JWT → Runtime validation → OAuth2 M2M to Gateway with Cedar policy enforcement, including a Pre-Token Lambda for identity propagation
  • Agent SDK agnostic: Ships with Strands and LangGraph pattern examples, claims to work with others
  • Local dev environment: Docker Compose stack for development without deploying
  • Steering docs for AI assistants: .amazonq/ rules and extensive markdown guides designed as context for coding assistants

Caveats

  • The “production-ready” claim in the description sits awkwardly next to the README’s insistence that the baseline is “very, very simple” and intentionally minimal — you’ll be doing real work post-deployment
  • “Vibe-codability” as a primary tenet is a new coinage; your mileage with AI assistants actually understanding Cedar policy syntax may vary
  • 508 stars suggests early traction but limited battle-testing

Verdict

Worth a look if you’re already committed to Bedrock AgentCore and want to skip the two-week infrastructure slog. Skip it if you’re not on AWS, not using AgentCore, or prefer to understand your auth flows rather than delegate them to an AI assistant’s interpretation of markdown docs.

heatdrop uses Google Analytics to see which pages get read — nothing else. Your call. How we handle data.