2025 was about AI agents. They were everywhere, everyone wanted to get an AI agent into their product. And software development, of course, was not an exception. Unless you were living under the rock, you've at least seen how AI agents could help developers become more productive as LLMs got tools and suddenly were able to go beyond just text output. They got access to file systems, code versioning systems, databases and more. It didn't go well in all cases, some databases got wiped, a few production systems got (a bit of) downtime, but otherwise this was a big step up in quality of LLM answers and productivity improvement.
But why stop at a single AI agent, when you can have several? You can make them talk to each other, work with each other, compete with each other, on the premise of getting much better results. In this talk we'll go through multi-agent systems from reasons for having multiple agents collaborate with each other to actual live code example of multi-agent system implementation to ways you can start using those systems in your everyday life.