Building an A2A Agent for telex.im using Mastra

The Challenge

For my HNGi13 Stage 3 backend task, I needed to create an Agent-to-Agent (A2A) integration with Telex.IM using JavaScript/TypeScript. I had to use Mastra AI which was both exciting and a bit challenging since I was new to building AI agents.
I built a book suggestion agent that suggested 5 books to a user depending on the topic they submit.

Getting Started

Learning the Ropes with Mastra AI

I began by setting up a Mastra project, which came with a weather agent example. This was super helpful because it showed me how everything fits together. The example covered:

• How to set up an agent
• Adding tools and workflows
• Handling API requests

Building My Book Suggestion Agent

Using the weather agent as a guide, I built my book suggestion agent. Here's how it went:

1. Setting Up: I organized the code and set up TypeScript to catch my mistakes early.
2. The Book Tool: Created a tool that fetches book recommendations from Open Library.
3. Conversation Flow: Made sure the agent could handle different types of requests.
4. Error Handling: Added friendly messages for when things don't go as planned.

The Tricky Part: Telex.IM Integration

This is where things got challenging. I struggled with:

• Connecting my A2A agent to telex from my hosted mastra site
• Getting the message format just right - had some issues with the string but I figured it out
• making requests was initially slow and also giving results I did not want but with a couple of code fixes and more deploys I managed to make it work.

The HNG team provided a helpful blog post that explained how to connect Mastra with Telex.IM which made things much more easier.

What I Learned

The best part? It worked! Now I can:

• Get book recommendations through Telex.IM
• See how people interact with my agent
• Fix issues when they come up

You can check out the interaction logs here to see it in action.

My Big Takeaways

1. Examples are gold: The weather agent was my best teacher.
2. Ask for help: Other interns saved me hours of frustration.
3. Keep trying: The solution often comes right after you want to give up.
4. Use what you're given: The HNG team's resources were a lifesaver.

This project was a huge learning experience. I went from knowing nothing about AI agents to building something that actually works. These skills will be super useful in any agent I work on next!