Add Authoritative Context to Your Agent

​Learn how to make your agent more reliable by using processors to supply current, authoritative context.

​A team member changes an SQL query. Later, your agent replaces it with an old version from the conversation and reports success. No error occurs. The query still runs, and the results seem correct. You find the problem days later, when two people use the same query but report different figures.

​The agent created the query from conversation history instead of reading the current version. As a result, it could not see changes from another session, a background process, or another agent. More memory can make this problem worse.

​In this workshop, Mora’s co-founders show a more reliable method. Their system reads the current query when it receives a request and gives that query to the agent as input.

​The presenters will show how to build this method in Mastra. You will learn how to use an input processor, RequestContext, and Studio traces to control context and see the exact input that each model receives.

​You will leave with a practical method for agents that change shared data—and for agents that remove recent changes without explanation.

​Hosted by:

​Brandon Barros, Founding Field Engineer @ Mastra

​Xavier Pladevall, Co-founder @ Mora

​Eduardo Portet, Co-founder @ Mora

​Recording and code examples will be available to everyone who registers.