Welcome to the Mastra Changelog.
We’ll be publishing these every week, but since this is our first one, we’re covering ground from the last two weeks.
Let's dive in.
Suspend/resume in workflows
We’ve made it easier to build checkpoints, human-in-the-loop intervention, and/or monitoring into any of your workflows. You can simply call suspend inside any step (or between steps) in a workflow and resume it later.
State is stored in the Postgres DB in the mastra engine.
The mental model here is to think about suspend as an exception you're throwing, where the workflow will handle catching that exception and persisting state.
mastra create for new projects
You can now build projects in a standalone way using npm create mastra.
This will create a new project scaffold with directories and (if you want) examples. We've found this to be a cleaner DX for new users.
While you can still initialize within an existing project (mastra init), create generates a ready-to-use project scaffold—complete with recommended directories, configuration, and boilerplate.
mastra dev for an interactive local dev server
Here's Shane Thomas demoing Mastra Dev, a local development server for Mastra projects that helps you see and interact with all the agents, workflows, and tools within your Mastra application.
It includes:
- an agents playground so you can chat with your agents
/textand/generateendpoints for each of your agents so you can test them viacurlor call them over the network- visual workflow diagrams for each workflow
- endpoints for each of your workflows so you can test them via
curlor call them over the network - a registry of all of your tools
mastra dev makes it easier to build, understand, and debug your agents and workflows.
Mastra Examples: basic building blocks for anyone to get started
To make it easy for anyone to start building projects with Mastra, we released a collection of short examples of all the basic AI engineering Mastra primitives: things like generating and streaming text, describing images, calling a workflow, chunking text, and even implementing basic RAG (retrieval-automated generation) systems.
Each example links to a subdirectory of a Github repo, so you can npm install and run them locally.
API endpoint improvements
Previously, Mastra offered a single "generate" API for text responses with two flags for structured output and streaming.
We realized having four combinations of flags was confusing, so we've split generate into two separate APIs: generate and stream, each with a flag for structured output.