Module Summary

  • Goal: Use your SWE agent to build and submit a complete AI-generated application
  • Estimated Time: 15–20 minutes
  • Prerequisites: Completed Modules 1 & 2 with a working frontend and Coding Agent

🏁 Build and Submit → AI Only

The apps-by-agents challenge showcases what autonomous agents can build - from prompt to production.
Your job is to prompt your SWE Coding Agent (via the xpander frontend or your own). The agent must do everything else.

🧠 Code, branch, commits, and PR must be authored by the agent.
Humans can only send prompts - no manual edits allowed.

To keep submissions simple and compatible, only static apps using HTML, CSS, and JavaScript are accepted.
All submissions are automatically hosted via S3 + CloudFront and published in the Apps Gallery.


🧠 Prompt Examples

Feel free to be creative - your prompts can be anything, as long as the result is a functional, agent-generated app.
The examples below are just to spark ideas:


🚀 Agent Submission Flow

Your agent should complete the following:

  1. Create a branch using this format:

    may-2025/your-github-username/your-app-name
  2. Generate the app in a matching folder (e.g., may-2025/janedoe/pomodoro-timer/)

  3. Open a pull request that includes:

    • App description
    • Tools and models used
    • Screenshot or live preview (optional)
    • Confirmation the app was AI-generated
  4. You (the human) should review the pull request to ensure it’s complete and valid. Browse all open submissions here: 👉 xpander-ai/apps-by-agents - Pull Requests

🛑 The pull request and all commits must come from the agent.

Once merged, the app is deployed automatically to the public showcase using S3 + CloudFront.


✅ Submission Checklist

Make sure your submission meets these requirements:

  • Built and submitted by the agent (no manual edits)
  • Correct folder and branch structure
  • App opens and works in a browser
  • PR includes description and preview (if possible)
  • The app is functional, original, and clearly AI-generated
  • PR was reviewed by the human and meets all guidelines

🎁 Go Further in Module 4

→ Want to run your agent locally or change how it thinks?

Check out Module 4: Advanced Customization