Mastra 博客 Examples 专栏：33 篇 AI 智能体实战教程

33 posts on Examples from the Mastra blog.

How to Build an AI Software Factory with AI Agents in TypeScript

A governed, observable system where specialized agents own the repetitive work of software development, and humans own the judgment.

Give a Mastra agent persistent file storage with Archil

The @mastra/archil integration attaches an Archil disk to your agent's workspace — persistent file storage backed by your own S3 that an agent can read, write, run code on, search, and share, with files that outlive the run.

Add realtime voice to your Mastra agents with Inworld

Inworld's Realtime API now plugs into Mastra via @mastra/voice-inworld-realtime — one WebSocket session gives your agent speech in/out, semantic turn detection, barge-in, and tool calling, attached to the agent you already have.

Giving a Mastra agent web access with Bright Data

How a Mastra agent uses Bright Data's webSearch and webFetch in its reasoning loop — what comes back from each tool, why memory turns search into research, and when to choose the typed @mastra/brightdata tools over the Web MCP.

Build a RAG agent with Mastra and Elasticsearch

How Elastic's reference build wires Mastra's agent, tool, and memory primitives onto Elasticsearch's vector store to answer questions over a corpus of 500 sci-fi movies.

Building a voice research agent with Mastra, Render, and AssemblyAI

How ravendr — Render and AssemblyAI's reference voice research agent — combines the AssemblyAI Voice Agent API, Render Workflows, and per-stage Mastra agents to answer in under 60 seconds without dead air.

A tutorial on building a movie recommendation agent with vector search and Observational Memory — all on a single MongoDB Atlas cluster.

Automating multi-language PR descriptions in GitHub Actions

Build a GitHub Action that uses Mastra to read pull request diffs and generate multi-language PR descriptions.

Build a research coordinator with supervisor agents

Build a research coordinator that orchestrates specialized agents to research topics and produce detailed reports.

Building a code review bot

Build an automated code review bot that uses workspace skills to apply coding standards and provide PR feedback.

Building a coding agent

Build and run an interactive coding agent with tool approvals, persistent sessions, and a terminal UI.

Building a dev assistant

Build a development assistant that uses workspace filesystems, sandboxes, skills, and search to write and test code.

Building a docs manager

Build an AI documentation manager that creates, updates, and organizes Markdown files with a workspace filesystem.

Building a notes MCP server

Build an MCP server that lets clients list, read, write, and summarize a collection of Markdown notes.

Building a research paper assistant with RAG

Build an AI research assistant that analyzes academic papers and answers questions using retrieval-augmented generation.

Building a signal provider

Build a polling signal provider that watches an external service and pushes notifications into an agent thread.

Building an AI recruiter

Build a recruiter workflow in Mastra that gathers candidate information, branches on candidate profiles, and generates follow-up questions.

Learn how to migrate from Mastra’s VNext streaming methods to the new standard APIs, with details on renaming, compatibility, and key differences between AI SDK v4 and v5.

Building low-latency guardrails to secure your agents

How we built a suite of out-of-the-box input processors and optimized them from 5000ms to under 500ms per request.

Building an AI Research Assistant with vNext Workflows

A deep dive into building MastraManus, an AI research assistant powered by Mastra's vNext workflows.

Build a full-stack TypeScript agent with Mastra and Copilot Kit.

How we built an interactive detective game using Mastra's agent system and Next.js.

FlashGenius: instantly generate flashcards on any topic

How we vibecoded Flash Genius, an instant flashcard generator, in 1 hour

Building a Web Browsing Agent with Mastra and Stagehand

See how we built a web browsing agent using Mastra and Stagehand. Learn about the tools needed to allow your agents to control a browser.

Mastra's dynamic agents provide a powerful way to handle context in AI applications, without exposing sensitive information to the LLM or relying on globals.

Building a Personal Assistant with Mastra and MCP

How we built a personal assistant that can manage emails, monitor GitHub activity, schedule social media posts, and more - all through a Telegram interface.

PDF-to-Blog: Giving documents a second life with Mastra and Mistral OCR

A technical guide showing how to build a PDF-to-blog converter using Mastra and Mistral OCR

Mastra Text-to-SQL: from natural language to database queries

Use Mastra's Text-to-SQL system to convert natural language to database queries

Integrating Mastra with Next.js: Two Approaches for AI-Powered Apps

A comprehensive guide to adding Mastra's agent capabilities to your Next.js applications, with flexible integration options for projects of any size

How we used Mastra and Gemini Flash 2.0 to build a natural language interface for code repositories

How we built a whiteboard image to Excalidraw converter using Mastra's multi-agent workflows, and what we learned about breaking down complex AI tasks into deterministic steps with validation loops.

A demo using Mastra to show how multiple AI agents can work together to plan travel itineraries.

Building a NotebookLM clone with an agent orchestrator

A technical guide showing how to build a NotebookLM clone that transforms documents into AI-generated podcasts using Mastra's agent orchestration, Pgvector vector storage, and LlamaIndex for content processing.

Ship better agents