Limits
Verified request ceilings and where to find workspace-specific entitlements.
Your workspace's active subscription or installed license determines its entitlements. Check the in-app plan and usage display for your current allowances. The old Free/Pro quota table was not backed by the product and has been removed.
Technical ceilings
These are current implementation limits, not a pricing table or throughput guarantee. Deployment versions and additional governance checks can impose further restrictions.
| Operation | Ceiling |
|---|---|
POST /datamaker generation quantity | 100,000 per request |
| One plan run, totalled across its entities | 100,000 rows |
| Inline set contents | 10,000 rows and 5 MiB of serialized JSON |
| One dataset chunk append request | 10,000 rows and 5 MiB of JSONL |
| Dataset preview/sample response | Up to 1,000 rows |
| Configurable scenario timeout | 60–3,600 seconds |
| MCP write/destructive call | 10,000 rows and 5,000,000 payload bytes |
| MCP write requiring an explicit volume approval | More than 1,000 rows |
| Agent unmask request | 1,000 tokens per call |
| Audit event list | Default 50, maximum 200 per request |
A plan run has its own budget because a per-entity check does not bound one: ten entities of 100,000 rows each pass individually and still ask a single run for a million rows. Size the plan against the run ceiling, not only against the per-request one.
A dataset can exceed an individual chunk's limit. Use multiple chunks or a supported data job rather than placing a large dataset in one inline set. A migration-scale load has its own approval; a standard-load approval does not authorize it.
Usage and capacity
Technical limits do not describe AI credits, licensed team seats, configured core systems, storage capacity or worker concurrency. Those are separate settings and entitlements. See What is counted, AI usage and Core systems.
For self-hosted installations, capacity also depends on CPU, memory, disk, blob storage and runner configuration. Do not infer a log-retention period, overage price or hourly generation quota from the request limits above.
When a request is rejected
Read the status and response body. Split oversized payloads, fix invalid quantities, or ask the administrator to resolve the indicated license or policy restriction. Retrying the same rejected payload repeatedly will not increase its limit.