mastra-ai/mastra

Mastra is a TypeScript framework for building AI-powered applications and agents. It is designed to take developers from early prototypes to production-ready AI products, with all the infrastructure needed bundled into one framework. An "agent" in this context is a program that uses a large language model (LLM) to reason through a goal, decide which tools to use, and iterate until it produces a final result, without you having to script every step. Mastra makes building these agents easier by providing model routing (connect to over 40 different AI providers including OpenAI, Anthropic, and Gemini through one consistent interface), memory (so agents remember past conversations and can recall relevant information), tool integration, and workflow orchestration for multi-step processes. The workflow system lets you chain multiple steps, add branching logic, and even pause execution to wait for a human to approve something before continuing. This last feature, human-in-the-loop, is important for production use cases where some decisions need a person's oversight. Mastra integrates with web frameworks like Next.js and supports the Model Context Protocol (MCP), a standard for exposing tools and resources to AI models. It also includes built-in evaluation tools for measuring and improving agent quality over time. You would use Mastra when building a TypeScript-based application that needs AI capabilities, chatbots, autonomous research agents, workflow automation with AI decision-making, or any system where you want structured control over how an LLM behaves in production.