Give Your Mastra Agent a Computer
Most agents hit a wall - they can call an API or query a database, but they can't actually work with files, run code, or act on the results. Give them a filesystem and a sandboxed execution environment, and suddenly a whole new class of problems becomes solvable.
But that opens up harder questions. How do you isolate users and tasks? Where does persistent state live? How much freedom can you safely give an agent without exposing your infrastructure?
In this workshop, we'll build agents with Mastra that can create and edit files, run shell commands, install dependencies, and use execution results to decide what to do next. We'll cover how sandboxes and filesystems work together, when to use local development versus remote execution, and what state should survive between runs.
You'll see practical patterns for handling uploaded files, generated artifacts, and agent-created outputs. We'll also dig into the architectural problems - isolating users and tasks, preventing state leaks, choosing where persistent files live, and safely giving agents enough freedom to be useful.
By the end, you'll have a mental model for designing agents that can work with files and execute arbitrary tasks safely, plus patterns you can adapt to coding agents, data analysis, document processing, and other agentic applications.
Hosted by
Alex Booker, Developer Experience, Mastra
Recording and code examples will be available to everyone who registers.