Build an AI Agent That Handles the Task You Keep Putting Off

Everyone has that one thing. The bill you keep forgetting to pay. The appointment you meant to schedule three weeks ago. The weekly report that takes 45 minutes of copying and pasting from the same five dashboards.

What if you could build an AI agent that just handles it?

That's what the Agents for Humans Hackathon is about. Pick a repetitive task, build an agent that takes care of it, and submit by September 14.

But let's get into the actual building part.

What is Strands Agents SDK?
Strands is an open source Python SDK for building AI agents. It's production-ready and built by teams at AWS. The core idea: you define an agent with a model, a set of tools, and a prompt. The agent figures out how to use those tools to accomplish a goal.

For a step-by-step walkthrough on building your first agent with Strands, check out this guide: https://strandsagents.com/blog/an-agent-for-the-thing-you-keep-putting-off/

That's the skeleton. You define tools as Python functions, hand them to the agent, and it decides when and how to use them based on the task you give it.

What Could You Build?
Here are some real ideas across the three hackathon tracks:

Everyday Agents

• An agent that monitors your inbox for bills, extracts due dates, and sends you a digest every Monday
• A grocery agent that checks your calendar, sees you're hosting dinner Friday, and builds a shopping list based on the recipe you pinned

• A scheduling agent that coordinates availability across your family's calendars
  Professional Agents

• An agent that pulls data from your dashboards every week and drafts your status report

• A meeting prep agent that reads the agenda, pulls relevant docs, and gives you a 2-minute briefing before each call

• A code review agent that checks PRs against your team's style guide and flags issues before a human reviewer sees them
  Good Neighbor Agents

• A volunteer coordination agent for a food bank that matches available volunteers to open shifts

• An agent that helps a small nonprofit track grant deadlines and draft application outlines

• A school club agent that handles event logistics (room booking, email reminders, attendance tracking)

Getting Started in 20 Minutes

1. Install Strands - Strands Quick Start
2. Define your tools. What does your agent need to interact with? APIs, databases, file systems, calendars? Each tool is a Python function with a @tool decorator.
3. Create your agent. Pass in your tools and give it a task.
4. Test it. Run it locally, see what decisions the agent makes, and iterate on your tools and prompts.
5. Deploy (optional but recommended). You can deploy with AWS AgentCore for extra points on Technical Implementation, but it's not required.

What Makes a Strong Submission
Looking at the judging criteria, here's what stands out:

• Solve a real problem. The more specific, the better. "Helps with productivity" is vague. "Drafts my weekly client update email by pulling data from Jira and Slack" is specific.

• ** Show it working.** A 5-minute demo video is required. Screen recording with voiceover is fine. No need to be on camera.

• Clean code repo. Public repo with setup instructions so judges can actually run it.

• Write it up. Explain what it does, who it's for, and how it works.

• Live demo link. Not required, but projects with one score higher.

• Bonus: publish a post on AWS Builder Center about your build journey with "Agents for Humans" in the title for extra points.

The Details

• Deadline: September 14, 2026 at 5pm PT
• Prizes: $40K in cash across three tracks (Grand Prize: $10K)
• Credits: $50 in AWS credits for every participant who submits
• Teams: Solo or up to 5 people
• Open worldwide (some country restrictions, check the official rules) Register and submit: agentsforhumans.devpost.com