How to build Agents that act on Sensitive Data with Jev, Arcade, and Mastra

​Most agent demos stop at the point where the agent is about to do something real. It has reasoned well, it has picked the right tool, and then the video ends before it touches the production system, because that is the part nobody has figured out how to show safely. This event is about that part.

​Spend a day with the Mastra and Arcade teams building agents that actually act inside real systems, with the permissions, policy, and audit trail an enterprise would need before letting them anywhere near production. Mastra gives you the agent framework: workflows, memory, tool calling, and evals in TypeScript. Arcade gives you the actions runtime: authenticated access to the tools your agent needs (Gmail, Slack, GitHub, Salesforce, and a few hundred others), scoped to the user the agent is acting for, with policy hooks that decide what the agent may do before it does it.

​Who should come

​Developers building agents that need to touch production systems, and the people who have to sign off on letting them. If you have ever had a security or platform team ask "what stops it from doing something dumb," this is the day you get a real answer.

​What to bring

​A laptop, Node 20 or later, and a GitHub account. Everything else (API keys, sandboxed accounts, credits) is provided at the door. No prior Mastra or Arcade experience needed, though the Mastra quickstart and Arcade's Mastra integration guide will have you oriented in about twenty minutes if you want a head start.

​Food and beverages are on us, all day.