Start here
Economics and quotas
Creator seats, platform capacity, metering, quota behavior, and bring-your-own AI costs.
v1 design baseline. This page specifies intended behavior. Delivery and validation are tracked in the implementation plan; it is not a claim that the platform is already implemented.
Commercial structure
Annual platform subscription
+ creator/developer seats
+ dedicated runtime package
+ included usage
+ controlled overages/additional capacity
Application consumers are generally free.
An organization app used by 4,000 employees should not require 4,000 intrnl creator seats.
Commercial identities
App Consumer
- uses deployed apps
- no paid creator seat
Creator
- creates/edits through AI or console
Developer
- Git/API/advanced diagnostics
Admin
- governance and security
Packaging can initially collapse Creator and Developer if simplicity matters, while preserving the permission distinction internally.
Metering dimensions
Track:
- Active applications
- Preview environments
- Creator/developer seats
- Build minutes
- Concurrent builds
- Runtime requests
- CPU time
- Egress
- Database storage
- Snapshot storage
- KV storage/operations
- Artifact storage
- Log retention
- Dedicated cluster capacity
Quotas
Organization-level:
maximum creators
maximum active apps
maximum active previews
build minutes/month
concurrent builds
total DB storage
total snapshot storage
runtime request pool
egress allowance
log retention
Application-level:
runtime memory/CPU profile
DB size
KV size
request rate
body-size limits
outbound hostnames
log volume
preview lifetime
Quota behavior
Default enterprise behavior:
- Alert at 75%.
- Alert at 90%.
- Alert and prevent new optional consumption at 100%.
- Do not unexpectedly shut down production apps because of a routine usage overage.
- Throttle abusive or dangerous traffic immediately.
- Require explicit opt-in for automatic billable overages.
AI costs
For BYO AI clients:
AI inference bill
→ Organization/user AI agreement
intrnl bill
→ platform, runtime, security, storage, governance
A future embedded intrnl AI becomes a separately metered add-on.