
Mastra
Mastra is a TypeScript-first AI framework that provides agent primitives, durable workflows, RAG tooling, and observability for production AI applications. It includes a developer SDK, a CLI for local development, and a cloud product for deployment and monitoring. Mastra unifies provider APIs and supports memory, tool calling, and vector-backed retrieval to equip agents with real-world data and long-term context.
- TypeScript agent framework — build agents with first-class TypeScript types and primitives to define instructions, memory, tools, and workflows.
- Durable workflows — author graph-based, durable state machines with clear step/branch semantics, suspend/resume, and real-time watch streams for long-running processes.
- Unified provider API & Model Router — swap LLM providers with a single API change and access many models through a unified router with IDE autocomplete.
- Memory & RAG — persist long-term memory, sync SaaS data sources, upsert and query unified vector stores, and use metadata filtering and agentic RAG for context-aware responses.
- Tool calling & orchestration — allow agents to call functions, trigger external actions, and embed workflows as tools for complex automation.
- Observability (.ops) — trace every workflow and agent step, capture performance metrics, run evals, and emit OpenTelemetry traces for debugging and monitoring.
To get started, install the SDK via the provided npm initializer, run the CLI for local development, and follow the docs to create agents, wire memory and tools, and deploy to Mastra Cloud for observability and scaling.
No discussions yet
Be the first to start a discussion about Mastra
Demo Video for Mastra
Developer
Pricing and Plans
Open-source SDK
Open-source Mastra SDK and examples available on GitHub for local development and experimentation.
- TypeScript SDK and CLI
- Agent and workflow primitives
- Examples, templates, and community repo
- Local development and testing