TanStack AI 1.0 is officially stable! 🎉 Build chat, voice, image generation, agents, and harnesses into your apps, with your choice of models, UI framework, and infrastructure. Your AI, your way.

A lot has happened since the alpha, and Alem, Jack, and Tom have put an incredible amount of work into getting here. Here's what's in 1.0, how to upgrade, and one more thing we've been working on 👇

That one more thing is an agent harness. Build an agent that keeps its session between requests, takes new instructions while it works, and runs in your terminal, web app, or editor. The harness APIs are experimental, and we'd love for you to try them.

Already using TanStack AI? Upgrade your stable packages together and follow the migration guide. The experimental harness packages stay on 0.x. Have fun building. 🌴

Congrats @AlemTuzlak @jherr @tombeckenham and @tannerlinsley ! 🚀

Congrats @tannerlinsley and team! What a special achievement!

tanstack team is killing it!

huge milestone for tanner and the crew

very excited for this release, and happy to adopted tanstack so early for our product stack.

All rights I didn't expect this

woah

TanStack AI stable release gives builders real freedom of choice

the ai helper in a school platform i built already switched models once so i dont hard wire client apps to one provider anymore Congrats Tanner 🎉

Steering the harness while it works is the bit I'd test first. Does a new instruction land between tool calls, or can it cut into one that's already running?

congratulations on the release! Can't wait to see what creators build with this.

nice that the harness bridge now refuses needsApproval tools. in 0.x they just ran unapproved

congrats team.

It's been a year since I saw the alpha version. Huge contract man @tannerlinsley !

Nice, currently building a service using it to learn more and how it is different from Langchain. 👍🏻🧑🏻‍💻

Stable and feature complete. Let’s go! Way to go team. Huge amount of effort here!

in 1.0, does the agent harness log tool steps apart from the chat ui, or do you ship that as one package?

I already was using it ..it 's really great

TanStack AI 1.0, tried to dm you about an engineer shipping a multi-model ai chat app i can intro you to but your dms are closed

curious where the trust boundary sits once agents can call tools across those integrations?

"your ai, your way" is the right pitch. swapping models without rewriting the app is the thing i actually want

does the harness layer stay model agnostic when providers handle tool calls differently?

amazing stuff, i love the new chat primitives 🫡

Chat, voice, image generation, agents, and harnesses can all be built into apps with a choice of models, UI framework, and infrastructure.

Congrats on 1.0. Swapping the agent backend without touching the React side is the part I'm most excited to try.

The first thing I'd build with your harness is a model-swap test: replay saved tasks and compare tool actions, since matching answers can hide different behavior.

Your AI, your way with full chat, voice, image and agents TanStack AI 1.0 stable is a huge win for builders who want freedom!

Can’t wait to try this out

can replace litellm already?

Looking real cool!

Finally a flexible way to build AI features

The unsexy SDK layer ends up deciding who actually ships agents, way more than the model leaderboard does.

Tip for anyone shipping chat UIs on this: render tool calls as a skeleton sized from the tool's output schema while args stream in. The result then lands without the layout jumping, which is what makes agent UIs feel cheap.

model choice without rewriting the app is the abstraction developers actually need