Mastra

Mastra is used as an agent and workflow framework in Devman’s AI systems practice. Public documentation should focus on patterns, not private deployment details.

Agent Definitions

A Mastra agent should have a clear purpose, explicit instructions, model configuration, tools, and storage assumptions. Avoid treating the framework as a black box: name the agent’s responsibility and keep tool capabilities narrow.

Workflows and Tools

Use workflows when the task has durable steps, branching, or background work. Use tools for bounded operations such as fetching records, creating tickets, querying a knowledge base, or calling an external API.

Storage and Memory

Memory and storage should be chosen according to the use case. Local development can use lightweight stores, while production usually needs durable infrastructure, backups, and observability.

Framework Discipline

Prefer the documented Mastra CLI and configuration model. Avoid workarounds that hide framework lifecycle, bypass tool schemas, or make builds depend on runtime-only state.

Related pages: Tools and Workflows and Deployment Patterns.