Why & How: A2A & MCP | Making An Agent2Agent Server

I wanted to share what I've learned from building an A2A Server, Client and integrating it into my chat app. 1) How do the Model Context Protocol and the Agent2Agent protocol work together? Why do we need both? Do we actually need both of them? What do I think will happen to them in the future? 2) How would you go about integrating A2A into your application? If you already have an MCP client, do you replace it, keep them both, or combine them in some other way? 3) Along the way I've published two projects that you're free to use: 📦 Hono Agent2Agent Server (Typescript) https://github.com/iamarcel/a2a-server https://www.npmjs.com/package/hono-a2a-server Mostly based on Google's example server, but ported to Hono (who still uses Express in 2025?). 🤖 Deep Researcher Agent (Typescript) https://github.com/iamarcel/deep-research-agent OpenDeepResearcher implemented using that Hono server. An example of an Agent2Agent server you can use and deploy, or adjust! --- Thanks for watching! Don't forget to subscribe and leave a comment so I can make more things you like even more ;)