Mastra agents with memory full stack example
Co-founder, Trigger.dev
We've released a new example project demonstrating how to build an AI agent with memory persistence using Mastra and Trigger.dev.
Mastra is a Typescript agent framework for building AI agents. Mastra handles AI agent memory and coordination while Trigger.dev handles orchestration, reliability, observability and scaling.
The example lets you enter a city and activity to get personalized clothing recommendations based on real-time weather data. Under the hood, it demonstrates:
- OpenAI GPT-4 powered agents with specialized roles using Mastra's agent orchestration framework
- PostgreSQL memory persistence enabling data sharing between agents with Zod schema validation
- Durable task execution with Trigger.dev's retry logic, observability, and
triggerAndWaitorchestration - Serverless-ready architecture using Node.js with Open-Meteo API integration and custom tools
export const whatShouldIWearTodayTask = task({ id: "what-should-i-wear-today", maxDuration: 15, run: async (payload: { city: string; activity?: string }, { ctx }) => { const threadId = ctx.run.id; // Step 1: Weather agent stores data in shared memory const weatherResult = await weatherDataTask.triggerAndWait({ city: payload.city, }); // Step 2: Clothing agent reads from shared memory const clothingResult = await clothingAdviceTask.triggerAndWait({ city: payload.city, activity: payload.activity || "walking", threadId: weatherResult.output.threadId, }); return clothingResult.output; },});export const weatherDataTask = task({ id: "weather-data", run: async (payload: { city: string }, { ctx }) => { const weatherAnalyst = mastra.getAgent("weatherAnalyst"); const response = await weatherAnalyst.generate( `Get weather for ${payload.city} and store in working memory`, { threadId: ctx.run.id, resourceId: payload.city } ); return { threadId: ctx.run.id }; },});
Check out the full example in our GitHub examples repository to see how Mastra's agent orchestration combines with Trigger.dev's durability and observability for production-ready AI workflows.