Back to News
1

Feature Story

GitHub - mastra-ai/mastra: the TypeScript AI agent framework

Feb 19, 2025 · github.com
GitHub - mastra-ai/mastra: the TypeScript AI agent framework
Mastra is a Typescript framework designed to expedite the development of AI applications by providing essential primitives such as workflows, agents, retrieval-augmented generation (RAG), integrations, and evaluations. It supports local and serverless cloud deployment. Key features include LLM Models, which use the Vercel AI SDK for model routing across providers like OpenAI, Anthropic, and Google Gemini; Agents that enable LLM models to perform actions using tools and workflows; and Tools, which are typed functions with integration access and parameter validation. Workflows are graph-based state machines with features like error handling and OpenTelemetry tracing, while RAG allows for constructing knowledge bases. Integrations offer type-safe API clients for third-party services, and Evals provide automated tests for LLM outputs.

To get started with Mastra, you need Node.js (v20.0+) and an API key from an LLM provider such as OpenAI, Anthropic, or Google Gemini. The `create-mastra` CLI tool helps in setting up a new Mastra project, and running `mastra dev` opens the Mastra playground. For contributing, developers are encouraged to discuss their ideas by opening an issue before submitting a pull request. The project also has a community Discord for support and encourages users to star the project on its page.

Key takeaways

  • Mastra is a Typescript framework designed for building AI applications quickly, offering features like workflows, agents, RAG, integrations, and evals.
  • It supports LLM model routing through the Vercel AI SDK, allowing interaction with providers like OpenAI, Anthropic, and Google Gemini.
  • Mastra includes tools, workflows, and integrations that can be used by agents to perform tasks and access third-party services.
  • Users can contribute to the project through coding, testing, and feature specification, with support available via a community Discord.
View Full Article

Discussion (0)

Be the first to comment!