so WorkOS ($20M ARR auth provider for OpenAI, Plaid) went from "how do we demo AI agents?" to shipping production workflows in ~2 months one engineer built custom enrichment tools that take messy prospect data and turns it into neatly formatted tables another scrapes

⏀AG-UI CLI + Mastra Easily start building a Frontend for your Mastra agents with a simple command line. Get a simple starter repo with building blocks including: - Generative UI - Shared state (between your app & agent) - Frontend tool calls Get started 📟 npx

we just shipped mastra templates pre-built projects for common agent use cases like pdf chat, text-to-aql, research agents, csv analysis, browser automation, docs chatbots zero to prototype in seconds npx mastra@latest create --template [name]

hey friends, this month I'm working with Mastra to help them launch templates! there are many new templates to talk about, but my favourite is this text-sql one we built never write a sql query again! just query the db in natural language 🚀

Agent Memory Workshop x.com/i/broadcasts/1…

Gemini, ChatGPT, Grok, and Perplexity each have their own take on a deep search agent. In this video, I show you how to build your own using TypeScript, Exa, and the new Mastra deep search template - all in just 3 minutes 👇🏻

btw mastra playground now shows agent reasoning in real-time basically you can watch your agents "think" through problems step-by-step, as they work helpful for debugging/understanding reasoning models + agent behavior

The Mastra MCP Client class is one of the most feature rich plug-and-play implementations out there mastra.ai/en/reference/t… What a great foundation that is. Choosing Mastra as the basis for PerfAgent just keeps paying off!

Started Mastra 101 Mastra. Love how the course is structured and how AI agents are directing us for this course.

wdym you havent built a mastra template yet? we're teaching you tomorrow lu.ma/wav8tlek

mastra.ai/blog/building-… Some lessons learned while building out agent input processors for Mastra 1. Every token matters 2. Every token REALLY matters 3. 1 & 2 (P.S. we shipped input processors)

we shipped input processors

with huge context windows, you might think you can just stuff everything in you can't in practice, large language models still hit real limits. that's where context engineering comes in

love learning out about tech lore - and was today years old to realise Vercel has a Finnish co-founder 🇫🇮 Tony Kovanen was the CTO between 2015-2017 what was known as ZEIT, before rebranded into Vercel in 2020. also, co-authored Next.js - arguably one of the most popular

If you're building TS agents and haven't tried Mastra, you're flying blind. Game changing stuff. Shoutout to Abhi Aiyer and the team!

Today’s a big day! We have something really exciting to share with the open-source community. We’re launching two open-weight language models: gpt-oss-120b and gpt-oss-20b. They’re incredible models, built for developers, trained for reasoning, efficiency, and real-world use.🧵

how to build a local AI agent using the new OpenAI OSS models no api keys required!

Honestly, when OpenAI’s new models were announced, I didn’t fully get the hype. I wondered: - How good is GPT-OSS compared to proprietary models? - What does this shift mean for the industry? - Can I actually build a local agent with it? I made this video to explore those

introducing mastra scorers evals still work but we learned a lot from watching how people used them users want to be able to do things like use evals on workflow steps businesses need custom evals tailored to KPIs so we shipped a bunch of improvements and are calling it

wishing everyone who celebrates a very happy gpt-5 release day but especially so to anyone who ships in typescript 300 spots ⬥ one day ⬥ builders only ⬥ are you in? tsconf.ai