I'm working on an AI agent that uses LangGraph, GPT 4.1, Docker, and tool calling to build & modify Rails applications. It's fascinating how much you can accomplish with tool calling. This is part 1 of me going over the approach, including using multiple docker containers, mounted volumes, and tool calling to overwrite files within the mounted volumes, that can then render the changes in real time on an iframe. Part 2 coming soon.
Absolutely! Rails continues to power incredibly productive teams and large-scale apps (GitHub, Shopify, Airbnb, Stripe, etc.). Rails uniquely enables small teams—or even solo developers—to build powerful, integrated apps quickly, which is exactly the environment I want for an AI agent.
Rails’ structure (ActiveRecord, migrations, generators, rich conventions) provides a deeply introspectable, consistent, and predictable environment—perfect for autonomous AI agents to learn, understand, and reliably execute tasks.
Other frameworks often require more custom glue or boilerplate. Rails is "batteries-included," which lets an AI agent leverage existing integrations (like Twilio, Gmail, Stripe, Quickbooks, etc.) out-of-the-box, making it the ideal playground for experimenting with powerful, generalized agents.
I built my first startup on Rails, it's extremely powerful and productive for solo devs, and my tool of choice for building quickly. :)