Skip to content

Tiers and current limits

Tier App CPU App RAM
Small 130m, approximately 1/8 vCPU 256 MiB
Medium 250m, 1/4 vCPU 512 MiB

These are the currently qualified tiers. Kata/VM overhead stays separate. Each app uses one writable pod, Service and retained data volume. Additional writable replicas and high availability need a replication/fencing design.

The current default app volume is 1 GiB. That physical capacity differs from usable payload. A prior snapshot fixture qualifies 896 MiB; it is not a blanket native-auth or full-workload capacity promise. Large native datasets and recovery still need separate qualification.

Resource Current contract
Account object allowance 10,000,000,000 bytes shared across apps
One File 2 MiB
File transfer per app 1 at a time
Unattached File 1-hour lifetime, then bounded cleanup
Fresh native user count No commercial app/owner count cap
Fresh native global sessions No small fixed count cap
Native session lifetime One hour absolute; refresh does not extend it
Password work One task per app with a bounded memory profile
Auth operations Separate budgets; retry 429 responses with delay

Removing count caps does not mean unlimited users or throughput. Database size, CPU, memory, cleanup, password concurrency and operation budgets still apply. Older explicit auth and SQLite File profiles retain their stored limits. Platform account registration also has a separate development quota.

DSL input is limited to 256 KiB. Generated project output is bounded before writes. The platform admits a finite number of builds and operations; overloaded requests can fail explicitly. The default CLI mutation wait is fifteen minutes. Inspect the accepted operation after a timeout instead of assuming failure.

Business requests and work histories also have payload/row bounds. Read events/jobs and the selected OpenAPI contract.

Full V1 implementation, arbitrary packages/network access, general schema migration, Apple login, cross-app account linking, a managed CORS policy for unrelated browser origins, automatic product restore, permanent purge and replacement-host recovery are not current promises.

The platform runs on a development node. A route working today does not establish a production uptime SLA, 100-app capacity, per-app disk fairness or full-account backup duration. Read the exact feature matrix.