Postgres infrastructure for AI platforms
Provision a production database when an agent builds an application, and create a production-like branch for every deployment preview.
Scale from the first user to millions of isolated databases and branches, in your cloud and close to the applications your platform creates.
A production database for every application. A branch for every preview.
AI platforms need databases at two different moments: when an application is created, and whenever that application is previewed or changed.
A production database for every app
When an agent builds a new application, Xata provisions an isolated Postgres database for it to use in production. The application receives a standard Postgres connection string and can start storing users, application data, agent state, and generated content immediately.
A branch for every deployment preview
When a preview deployment is created, Xata branches the application’s production database so the preview includes the same schema and production-like data. The agent can test code, migrations, and data changes together without affecting the live application or another preview.
Built for millions of databases and branches
AI platforms can create applications and previews continuously. Every new app may need a persistent production database, while every code change may create one or more temporary database branches.
Xata is designed to support this model at platform scale, from a few databases to millions of isolated databases and branches.
Provision on demand
Create databases and branches programmatically as users and agents start new workloads.
Isolate every workload
Give each application and preview its own schema, data, endpoint, and credentials.
Manage the complete lifecycle
Create, suspend, resume, and remove databases through your platform control plane.
Scale without shared staging
Run large numbers of applications, previews, and agent workflows without relying on a limited pool of shared databases.
Free-tier economics that scale
AI platforms can generate enormous numbers of databases and branches surprisingly quickly.
Imagine a platform with:
- 100,000 active users
- each user runs 5 agents or application workflows
- each workflow creates 10 previews, attempts, or deployment branches
That adds up to:
100,000 users × 5 workflows × 10 branches = 5 million database branches
Cost vs. branch count
Traditional provisioning vs. copy-on-write branching
With traditional infrastructure, millions of always-on databases and full database copies would make a free tier prohibitively expensive.
Xata combines scale-to-zero compute with copy-on-write branching, so costs follow actual activity rather than the total number of databases and branches your platform has created.
Scale to zero
Idle databases and branches stop consuming active compute while their schema and data remain available.
Copy-on-write branching
Preview branches share unchanged storage with production and store only the blocks that change instead of copying the full database.
Costs aligned with usage
Infrastructure spend is driven primarily by active compute and changed data, not by every free user, database, or preview that exists.
Offer a free tier without breaking the bank
Give every user a real Postgres database and every deployment preview an isolated branch, without being afraid that dormant free-tier workloads will consume always-on infrastructure.
Provision millions of databases and branches while paying primarily for the workloads that are active and the data that changes.
Deploy where your applications run
Choose Xata Cloud for a fully managed experience, or deploy Xata in your own cloud with BYOC.
Xata Cloud
Get started quickly with a fully managed Postgres platform operated by Xata. Provision databases and branches through your platform without managing the underlying infrastructure.
Bring your own cloud
Run Xata in your own cloud environment, close to the applications your platform creates.
BYOC helps you keep application and database traffic in the same cloud, reduce latency and data-transfer costs, meet residency requirements, and use existing cloud agreements.
Use Xata Cloud for simplicity, or BYOC when you need control over cloud, region, and infrastructure.
Built into your platform
Your control plane calls the Xata API or CLI to create, manage, and remove databases and branches as part of its normal application lifecycle.
Each application gets its own production database. Each preview, sandbox, or agent workflow gets an isolated branch created on demand and removed when the work is complete.
- App created
- production database provisioned
- credentials returned
- Preview starts
- branch created
- preview connects
- preview ends
- branch removed
White-label by design
Xata runs behind your platform. Users connect through standard Postgres drivers and connection strings, with no Xata-specific client or branding.
Controlled through your platform
Create, manage, suspend, and remove databases and branches programmatically through the Xata API, CLI, or your existing control plane.
A partnership designed for platforms
Platform pricing is based on capacity and deployment requirements rather than charging for every branch created.
This gives you predictable infrastructure costs as usage grows, without forcing you to limit how many databases, previews, or agent workflows your users can create.
Capacity-based pricing
Align pricing with the size and activity of your platform rather than raw branch count.
Dedicated co-engineering
Work directly with the Xata team on architecture, control-plane integration, deployment, and scaling.
Cloud deployment options
Choose Xata Cloud or BYOC based on your infrastructure, regions, and customer requirements.
Direct support
Use a shared Slack channel for integration support and ongoing architecture discussions.
Build your platform on Xata
Provision a production database for every application and an isolated branch for every preview, agent, and workflow.