Agents Within Agents Within Agents

Inception isn’t just an awesome movie; it’s a technical pattern I keep running into.

I mean I do love the movie, who can argue with such a great cast and incredible visuals. But I feel like I live through inception more than most. No not dreams within dreams, but that could be fun. The type of inceptions I live are all technology bound.

The first time I really felt like I was living through inception, and it continues to this day, was with Docker containers. If you aren’t careful you can definitely get trapped in limbo, unsure if you’re in the container, outside the container, if the container has access to the network, etc. Docker has gotten better and easier to understand and my understanding of it has also grown. Now it is a dream to work with and a vital part of many of my projects.

More recently, after taking a role working as an SE for Rasa, I find myself experiencing another form of inception: agents within agents within agents.

I was working on a tool to quickly create personable and personalized conversational agents (chatbots) from nothing more than a URL. I built an agent that takes the URL and orchestrates other agents, all in the name of… building an agent. In this case I was using Claude Code complete with Rasa’s MCP server and skills to build a Rasa agent powered by OpenAI. But when something isn’t working quite right I have to decide how to tweak the agent building the agents to build better agents.

Whew.

Ever feel like your work would make Leonardo DiCaprio’s dream within a dream proud? I’d love to hear about your version of inception!