I started building Mirgenta around May 2026. I saw a job posting for AI Agents and I wanted to build my own AI Agent. I read this blog post by Anthropic https://www.anthropic.com/engineering/building-effective-agents and one point that stuck with me was the question of why?
Why did we need an AI Agent System when a traditional software system could do the same thing?
Why did we need an AI Agent System when a traditional LLM could answer the same question?
Why did we need an AI Agent System when we could consult a human expert instead?
But the job posting described building AI Agents, so I figured that they must've had a good reason for building an AI Agent System. So I had to come up with a good reason myself.
I recall brainstorming with Claude for 3 hours on a Friday night, trying to come up with a good use case for an AI Agent System. I asked it to spitball some ideas and it would give the generic "Coding Agent" or "Research Agent". To be honest, none of those ideas really moved me. I felt like I was drowning in a red ocean.
Finally, a simple wish hit me: "I wish AI could listen to music with me and give me recommendations for similar pieces, or enjoy the melody along with me, or appreciate the stylistic intepretation that the performer utilized."
That is why I decided to build Mirgenta.