Shipping Production Agents with Non-Devs on Agent Builder
Getting AI agents into production forces a bad tradeoff. The people who know what an agent should do, like product, ops or marketing, can now build one in a no-code builder. But what they build lives in a UI that engineers can't review, test, version or deploy. So either the agent never ships, or developers rebuild it from scratch and the business can no longer change it. This workshop shows a way for both teams to work on the same agent.
In this live workshop, we'll build an agent in Mastra's Agent Builder without writing code, then connect it to a TypeScript codebase. The Builder and the code stay in sync at three points:
When you link the agent, the Builder turns it into a Mastra project in your repo, with its instructions, model, tools, memory and skills.
When someone edits the agent in the Builder, the change is committed to the repo automatically.
When developers add to the agent, their code goes in files the Builder never changes. Every sync shows up in git history.
You'll leave with a template you can fork. Swap in your own tools and point the sync at your repo. Your business team can keep improving the agent after engineering puts it in production.
Hosted by:
Brandon Barros, Founding Product Advocate @ Mastra