Mastra
: How to use it, features, and the business problems it solves

Last updated: 2026.08.09

# Code Generation # AI Model Development & Management # API & SDK # AI Agent

Add bookmark

What is Mastra?

An open-source AI agent development framework written in TypeScript. It integrates Agents/Tools/Workflows/RAG/Memory/Evals/Observability and lets you switch among LLMs like OpenAI, Anthropic, and Google Gemini. Under "build, observe, and improve," it suits building and running agents that run for days. The framework is free (OSS), with a hosted Mastra Platform available.

Business problems it solves

About "Mastra"

What is Mastra

Mastra is an open-source AI agent development framework you write in TypeScript. It bundles everything you need for agent development — Agents, Tools, Workflows (for complex logic), RAG (retrieval-augmented generation), Memory, Evals, and Observability. Under the banner "Build, observe, and improve agents," it has become a widely supported "leading TypeScript agent framework" suited to building and running agents that run for days (27k+ GitHub stars; used by Replit, SoftBank, Sanity, WorkOS, and Brex).

The Mastra top screen. The heading "Build AI agents," the line "Build, observe, and improve agents that run for days with Mastra, the leading TypeScript agent framework," the npm create mastra command, and on the right an Evals results table (Experiments / Scorers / Datasets) showing scores like correct tool selected 100% and incorrect tool used 62%

Source: Mastra official site

Type

Open-source AI agent development framework

Language

TypeScript / JavaScript

LLMs

OpenAI, Anthropic, Google Gemini, and more (switchable)

Key features

Agents / Tools / Workflows / RAG / Memory / Evals / Observability / Voice

Pricing

Framework free (OSS) + hosted Mastra Platform

Get started

npm create mastra

A complete set of agent-building blocks

Mastra's trait is that the pieces needed for agent development are integrated as a framework from the start, rather than cobbled together from separate libraries. You get TypeScript's type safety while covering everything from development through evaluation and operations.

01Build

Agents — Define agents combining LLMs, tools, and memory
Tools — Call external APIs and functions safely as tools
Workflows — Control complex logic with branching, loops, retries, and human-in-the-loop (HITL)

02Make it smarter

RAG — Add document retrieval to ground answers
Memory — Remember conversations and state to keep context
Voice — Add voice I/O for voice interfaces

03Observe and improve

Evals — Assess agent quality with Experiments, Scorers, and Datasets
Observability — Trace behaviour with Traces / Metrics / Logs
Multi-LLM — Switch the backend model with a config change

How to use

Create a project

Run the following in your terminal.
```
npm create mastra
```
Choose initial settings

Interactively select the project name, components to use (Agents / Tools / Workflows, etc.), and default LLM provider (OpenAI / Anthropic / Google Gemini, etc.).
Build an agent

Define an agent in TypeScript and wire in the tools and workflows you need.
Design workflows

Build workflows with branching, loops, retries, and human-in-the-loop, and control the agent's behaviour.
Deploy and operate

Deploy locally or to the cloud. With Mastra Platform (hosting), you can offload observability and scaling too.

Features

01Development

TypeScript-based all-in-one — Agents, Tools, RAG, and Workflows in one framework
Multiple LLM providers — Switch OpenAI / Anthropic / Google Gemini via config
Voice integration — Build agents with voice I/O

02Control

Flexible workflow control — Branching, loops, retries, human-in-the-loop (HITL)
Storage — Manage run history and data

03Operations

Evals & Observability — Quality evaluation and monitoring via Traces / Metrics / Logs
Mastra Platform — Cloud for deployment, hosting, and observability

Pricing

The Mastra framework itself is open source and free (self-hostable). On top of that, "Mastra Platform" provides cloud deployment and observability.

Plan	Monthly	Highlights
OSS framework	Free	Full features, self-hosted
Starter	$0	100K events (+$10/100K), 24 CPU hours, 15-day retention, unlimited users/deploys
Teams	$250	1M events (+$8/100K), 250 CPU hours, 6-month retention, multiple teams, SSO, SOC 2
Enterprise	Custom	For scale. RBAC, audit logs, SLAs, dedicated support

Used purely as a framework, it's free (OSS) — you can try every feature at no extra cost. Mastra Platform is the option when you want to deploy your agents to the cloud and run them with full observability (Traces/Metrics/Logs). Pricing may change, so confirm the latest on the official site.

Pricing is based on the official pricing page (USD) as of August 2026.

How it differs from other agent frameworks

In a space long dominated by Python frameworks (LangChain, LlamaIndex), Mastra stands out by centring on TypeScript/JavaScript and integrating evaluation and observability from the start. It suits teams building agents in the web/Node.js ecosystem.

Aspect	Mastra	Other agent-dev options
Language	TypeScript / JavaScript	Often Python
Scope	Integrates Agents–RAG–Evals–Observability	Often assembled from parts
Delivery	OSS + Mastra Platform (cloud)	Mainly libraries
Best for	Developers building with types in web/Node.js	Teams on an existing Python stack

To build AI apps/agents no-code, Dify; for business workflow automation, n8n is nearby (LangChain and LlamaIndex are Python-centric). If you want to build agents in code with TypeScript and keep evaluation and operations consistent, Mastra is a strong choice.

FAQ

Q. Is Mastra free to use?

Yes. The framework itself is open source and free (self-hostable). The cloud Mastra Platform has a free Starter plan; for heavier use, Teams ($250/mo) or Enterprise.

Q. What language do I use?

TypeScript / JavaScript. It plugs straight into the web/Node.js ecosystem.

Q. Which LLMs are supported?

Multiple providers including OpenAI, Anthropic, and Google Gemini, switchable with a config change.

Q. What can it do?

It covers what you need for agent development — Agents, tools, workflows, RAG, Memory, Evals, Observability, and voice.

Q. How is it different from LangChain?

LangChain is a Python-centric framework. Mastra centres on TypeScript/JavaScript and integrates evaluation and observability, making it suited to development in web/Node.js environments.

Information on this page is based on what was published on the official site as of August 2026. Features and pricing may change, so check the official site for the latest.

Try Mastra