Blog post tanstack.com/blog/tanstac...

We're targeting 1.0 for December 2025.

Try it and share feedback in Discord or GitHub!

Designed for your existing REST, GraphQL, or tRPC APIs. No backend migration.

Bonus: Works with sync engines (@electric-sql.com, Trailbase, PowerSync) for real-time updates with near-zero incremental network cost.

First query: ~100ms with targeted network request
Meanwhile: Full dataset syncs in background After sync: ALL queries run in <1ms client-side

Fast first paint + instant everything else.

Multiple components, same query? ONE network request.

Complex joins? Minimal batched requests.

Already-loaded rows? Reused automatically.

Fewer total requests than custom APIs, with better cache utilization.

Differential dataflow recomputes only what changed.

Mark a todo complete? Query results update in <1ms—even with 100k+ rows in memory.

No jitter. No loading states. Just instant updates.

0.5 adds 3 sync modes for different use cases:

- Eager: Load everything upfront (<10k rows)
- On-demand: Load only what queries need (>50k rows, search)
- Progressive: Load subset now, sync full dataset in background (collaborative apps)

This query:

useLiveQuery(q =>
q.from({ todos })
.where(eq(status, 'active'))
)

Automatically becomes:

GET /api/todos?status=active

No backend changes needed. Your queries become the API.

React made components pure functions: UI = f(state)

TanStack DB brings the same philosophy to data: view = query(collections)

You describe what data you need. DB handles fetching, caching, and updating—even across massive datasets.

... and new ones, like @arnoud.dev, who maintains the @tanstack.com angular adapter 🚀

Way ahead of you

The stack I just used

- @tanstack.com router in static mode
- @content-collections gives you a great api to local content + lunr.js to search
- MDX + import.meta.glob = 🐐. Every other way to do mdx is a black hole.
- @shiki.style for all syntax renders

😉