Add this to your agent's context
Give your AI agent the context it needs to work with Xata for platforms.
Developer platforms, SaaS products, and AI pipelines all share the same underlying need: a way to provision isolated Postgres databases programmatically, at any volume, without infrastructure costs scaling linearly with usage. Traditional database provisioning was designed for humans creating a handful of environments. At platform scale, thousands of preview environments, per-customer sandboxes, or per-agent databases, the economics break down. Xata's copy-on-write branching and scale-to-zero compute make per-database isolation viable at any volume.
The platform database problem
At platform scale, traditional database provisioning breaks down in three ways simultaneously: cost, utilization, and isolation. Each problem compounds the others.
Per-user cost scales linearly
Every preview environment, every customer sandbox, every developer branch is a full Postgres instance. At 10,000+ users, infrastructure cost dominates your margin. The math is straightforward: N users means N databases, each with its own storage footprint and compute allocation. There is no shared-storage efficiency, no tiering. Just linear cost growth with your user base.
Most databases sit idle
Preview environments are active for minutes per day. CI branches run for seconds per pipeline. Most per-customer sandbox databases are touched only occasionally. With always-on compute, you pay 24 hours a day for databases that are used 1% of the time. At scale, idle compute becomes the dominant cost, not actual usage.
Multi-tenancy is either expensive or risky
Shared databases solve the cost problem but introduce noisy-neighbor performance issues and compliance risk. Dedicated instances provide full isolation but bring the linear cost problem back. There has historically been no middle path that delivers true isolation at shared-database economics.
Copy-on-write changes the economics
Xata branches use copy-on-writestorage: a new branch shares the source database's storage until it writes something new. Creating 10,000 branches of a 1 TB database does not cost 10,000 TB. It costs 1 TB plus the incremental writes each branch makes. Combined with scale-to-zero compute, the marginal cost of an additional database approaches zero.
Cost vs. branch count
Traditional provisioning vs. copy-on-write branching
At 10,000 branches — marginal cost approaches zero
No per-branch fees
At platform scale, per-branch pricing would quickly become the dominant line item. Xata charges based on managed cloud spend, the underlying infrastructure cost, not branch count. Whether your platform runs 100 or 100,000 branches, the branch count itself does not drive the bill. You pay for compute time and storage writes, both of which scale with actual usage.
Built-in data isolation
Each branch is a fully independent Postgres instance with its own connection string. There are no noisy-neighbor effects between branches. A heavy workload on one branch does not affect query latency on another. Isolation is enforced at the storage layer, not the application layer, which means it cannot be misconfigured away.
BYOC and multi-cloud
Run the Xata data plane in your own cloud account on AWS, GCP, Azure, or Hetzner, and use your negotiated cloud discounts. Data stays in your jurisdiction, which matters for compliance and data residency requirements. Choose the region closest to your users with no vendor lock-in to a single cloud provider.
How platforms use Xata
Integration is API-driven. Your platform control plane calls the Xata API or CLI to create, manage, and tear down branches as part of its normal lifecycle operations, the same way it might call a cloud provider API to provision other infrastructure. Each customer, preview environment, or sandbox gets its own branch. Branches are created on demand and destroyed when no longer needed. Scale-to-zero means that forgotten or long-lived idle branches accumulate no compute cost.
Xata runs entirely behind the scenes. Your users see your platform, your branding, and your experience. The Postgres infrastructure is white-label by design: there is no Xata branding in connection strings, no Xata-specific client required, no changes to how your users interact with their databases.
Designed for platform volume
Developer platforms creating preview environments per pull request typically run 10,000–50,000 branches per day, with pricing based on capacity tier. SaaS products with dedicated per-customer databases run 1,000–10,000 concurrent branches, priced on managed cloud spend. AI and ML pipelines that provision an isolated database per agent execution, training run, or simulation pay only for active compute. Branches that are idle between runs cost nothing.
Partnership model
Xata's pricing for platforms is capacity-based: a predictable monthly cost tied to your scale tier, not per-branch metering. This means no billing surprises as your platform grows and no incentive for Xata to limit your branch count.
Platform integrations come with dedicated co-engineering support: a shared Slack channel, joint architecture planning, and integration assistance for connecting the Xata API into your control plane. See pricing details.