Xata - Postgres at Scale

@xata.io

With copy-on-write branching, data masking, separated storage & compute , 100% Postgres

A coding agent wants its own database for every task, created when the work starts and removed when it ends. A shared staging database doesn't survive that for long. Our CEO Monica Sarbu on what to do instead, today 2:50pm PT, Stage 2, WeAreDevelopers San José: "Databases in the Agent Era."

Need production data to debug a slow query? Normally, you'd have to isolate & validate data & deploy the fix. With the Xata MCP server and Codex, you just talk to an agent. Branch your Postgres database in seconds regardless of the size and let AI do the debugging. Watch 👇

Xata is coming to Azure. Branch-native Postgres. A copy-on-write branch of an existing database in seconds, whether the parent is 1GB or 1TB. Xata branches only incur costs for changed data plus compute while awake. A sleeping branch is basically free. Join the waitlist: xata.io/get/azure

pgp_sym_decrypt with the wrong key throws "Wrong key or corrupt data". No garbage bytes returned. The PGP integrity check earns its keep when every tenant holds a different key: the wrong key fails the query instead of quietly returning nonsense. Read about it here: xata.io/blog/multi-t...

Bild

Finding a slow query on a Postgres branch, without leaving the terminal. xata branch query-insights ranks that branch's statements by cost, flags the slow ones, and prints the command to inspect any of them. Logs and instance metrics are one command away.

If every agent and PR gets its own Postgres branch, you're scripting those, not clicking them. So the CLI output has to be parseable. Now it is: piped output is clean plain text, xata branch list --json for anything automated. Boring, but that's the point at scale.

2TB+ Postgres. Branch it in seconds, query it, delete it. xata scratch (new Xata CLI cmd) forks a disposable branch off your real DB, drops you into psql or runs SQL inline, tears it down on exit. Demo 👇

Introducing the Xata GitHub App. Connect a repo to your Xata project and every pull request gets its own isolated Postgres branch off main, with production-like data, ready in seconds when the PR opens and gone when it closes.

Xata is now on Google Cloud. Spin up an isolated Postgres branch from your production data in 3 seconds, in-region in us-central1 (Iowa). Every coding agent, PR, and CI run gets its own production-like database. No shared staging, no migration.

We run 100% vanilla Postgres. No fork, no modifications. A deliberate decision from day one to build a platform around Postgres, with branching at the storage layer instead of inside the database. This means the whole Postgres ecosystem just works with Xata.

Bild

Branching gives every engineer and every agent their own Postgres. Now every branch has its own logs. Read, filter, and search them in the console, or match a pattern in one click. Same logs in your terminal with xata branch logs.

Postgres CDC has always been a thing you babysit: a CLI in a screen session, secrets in a config file, a pipeline that breaks on the next ALTER TABLE. pgstream v1.1.0 is a step toward running it as a service. K8s probes, JSON logs, and schema changes that flow to the target. xata.io/blog/pgstrea...

pgstream v1.1.0: running Postgres replication as a service by Noémi Ványi

What's new in pgstream v1.1.0: env vars in YAML config, JSON logs, Kubernetes health probes, custom TLS for OpenSearch, and fixes for composite primary keys.

xata.io

Today, Xata CEO Monica Sarbu joined the Berlin Buzzwords panel "The Agent Era: How AI Agents Are Reshaping Data Platforms," alongside speakers from Snowflake, Elastic, and MotherDuck, on what breaks when agents become the most demanding users of data platforms.

BildBild

That sidebar full of "untitled query (3)" you never keep? We reworked the Xata query editor: new queries open as drafts, run and discard, nothing saved. Plus Run selected, per-statement tabs, autosave with AI naming. xata.io/blog/query-editor-v2-a-better-way-to-write-sql-in-xata-console

Query Editor v2: A Better Way to Write SQL in Xata Console by Marcello Novelli

We've rebuilt the query editor from the ground up. The new version is faster, smarter, and easier to use.

xata.io

We rebuilt how every Postgres branch on Xata is observed: out of a central observability vendor, into a per-cell VictoriaMetrics stack. Nothing crosses cells, so one cell's observability can't break another's. Six weeks of work, no user-visible downtime. Full write-up: xata.io/blog/how-we-...

How we rebuilt Postgres branch metrics on VictoriaMetrics by Alexis Rico

How we rebuilt Xata's PostgreSQL branch metrics on a self-hosted VictoriaMetrics stack in six weeks, with zero user-visible downtime.

xata.io

Monica + the Xata team are at AWS Summit Hamburg today, in the AWS Startup Zone, Hall 4. If you're thinking about Postgres for agent scale, swing by. Always up for that conversation.

Bild

pgstream v1.0.2 shipped. Stream-mode WAL batching means post-snapshot catch-up stops stalling on bulk-INSERT and DELETE tables (Postgres sink only for now; UPDATE still goes one at a time). Decode-time table filtering inside Postgres via wal2json.

Release v1.0.2 · xataio/pgstream

⚡ Highlights 🔀 Batch WAL operations in stream mode 🆔 New settings for wal2json plugin: add_tables and filter_tables 🐛 Bug fixes Dependency updates What's Changed 1dc7cc1 Add add_tables and filter...

github.com

Every Xata Postgres branch now has a managed PgBouncer endpoint, in the same memory envelope as the database. Auto-tuned to 0.9 of max_connections. 10,000 client connections per branch. Hibernates with the branch. Append `-pooler` to your branch ID. Writeup: xata.io/blog/a-conne...

Postgres connection pooling per branch | Xata by Noémi Ványi

Every Xata Postgres branch includes a built-in PgBouncer pooler in transaction mode. Tuned for serverless and edge, scales to zero with the branch.

xata.io