Module Summary
  • Goal: Learn how to work with your manager agent to build and submit an application using the web UI and CLI tools
  • Estimated Time: 10-20 minutes
  • Prerequisites: Manager agent with MCP and GitHub PR capabilities
🚀 In this final module, you’ll activate your agent fleet by working directly with your manager agent. You’ll use the web UI to instruct your manager agent, monitor its orchestration of developer agents, and submit your finished application to the xpander.ai repository. This is where your preparation in the previous modules comes together!

​
🕹️ Working with Your Manager Agent via the Web UI

The manager agent is your main point of contact for building applications with your agent fleet. You’ll interact with it through the xpander.ai web UI:
  1. Go to https://app.xpander.ai and go to your manager AI Agent (AI Development Manager)
  2. In the canvas, click the “Chat” source node and click the URL to open the Chat UI.
  3. Enter a prompt describing the application you want to build and instruct the manager agent to submit it when ready. For example:
example prompt
Build a minimal “QuickTodo” single-page app using vanilla HTML, CSS, and JavaScript. Users must be able to add, edit, and delete tasks; data should persist in localStorage so tasks remain after reload. Acceptance criteria: adding shows the new task immediately, editing updates it in place, deleting removes it, and a page reload restores all tasks. Code must pass ESLint. Target branch: may-2025/my-github-username/quick-todo
Change the my-github-username part to your GitHub username
The manager agent will interpret your request, coordinate the developer agents, and handle the submission process for you. Feel free to get creative! The manager agent will break down your request and assign tasks to the developer agents.
Your manager agent will coordinate the work between your developer agents, review their code, and select the best implementation. This all happens automatically thanks to the Agent-to-Agent (A2A) capabilities you configured.

​
✅ Confirm Agent Activity

After submitting your prompt in the web UI, wait a few moments for the manager agent to begin orchestrating the build process. You should see updates in the thread as the manager agent delegates tasks and coordinates the developer agents. You can also go to the Activity tab in your Manager AI Agent and see the progress.

​
What to Expect in the Logs

When you run xpander logs, you’ll see detailed output from your agents. Here are some examples and what they mean:
xpander logs output
Sandbox base directory: /usr/src/app/sandboxes
2025-04-30 11:17:10.616 | INFO     | xpander_utils.events.git_init:configure_git_credentials:17 - Setting up GIT for user dudutwizer
2025-04-30 11:17:10.621 | INFO     | xpander_utils.events.git_init:configure_git_credentials:29 - ✅ Git configured with PAT
2025-04-30 11:17:10.628 | INFO     | xpander_utils.events.listener:register:287 - Starting events listener
2025-04-30 11:17:12.713 | INFO     | xpander_utils.events.listener:_register_agent_worker:238 - Worker registered - 12687c2c-467f-4a38-8cbf-54efccfd2ed9
2025-04-30 11:17:12.713 | INFO     | xpander_utils.events.listener:_register_agent_worker:244 - Agent Name: dev-agent-1
 Chat URL: https://chocolate-jaguar.agents.xpander.ai
loading agent e62daf10-b8fd-45c8-8862-6168fb20b3c9
🪄 Starting Agent Loop
🧠 Thread id: 56dee347-dce3-42b3-b73f-0ff904ccf497
--------------------------------------------------------------------------------
🔍 Step 1
running tool xpfinish-agent-execution-finished on agent e62daf10-b8fd-45c8-8862-6168fb20b3c9 with execution e28f7897-988f-4450-b947-794bb2f17d94
✅ xpfinish-agent-execution-finished
After a few minutes, the manager agent will update the thread in the web UI with the results and next steps. You can follow along and interact as needed.

​
🏆 See What Everyone Built!

At the end of the workshop, we’ll showcase all the applications created by participants and their agent fleets. This is a great opportunity to:
  • See different approaches to agent orchestration
  • Explore creative applications built by AI agents
  • Get inspired for your next AI agent project

​
✅ Checkpoint

By completing this module, you should now be able to:
  1. Use MCP to communicate with your manager agent through Cursor
  2. Have your manager coordinate your developer agents to build an application
  3. Monitor agent activity and collaboration in real-time
  4. Create a GitHub PR with your application automatically
  5. Submit your creation to the workshop repository

​
🎓 Workshop Completion

Congratulations! By completing all five modules, you’ve:
  1. Set up a GitHub MCP Server to enhance your Cursor IDE
  2. Built your first developer agent
  3. Expanded your agent fleet with different models and capabilities
  4. Created a manager agent to orchestrate your development process
  5. Used your agent fleet to build and submit a real application
You now have a powerful understanding of how to build and deploy AI agent systems using xpander.ai’s Agent-to-Agent (A2A) framework, Model Context Protocol (MCP), and Amazon Bedrock.

​
📚 Additional Resources