> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://tieback.io/docs/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://tieback.io/docs/_mcp/server.

# Limits & Performance

> Platform ceilings, throughput benchmarks and scaling characteristics for minting, including over-mint protection and identity uniqueness guarantees.

## Platform Limits

| Limit                          | Value                         |
| ------------------------------ | ----------------------------- |
| **Max units per mint request** | 1,000,000                     |
| **Processing cycle size**      | 20,000 units                  |
| **Max lots per request**       | 100                           |
| **Lot quantity sum**           | Must equal requested quantity |
| **Export pack size**           | 10,000 units per CSV pack     |
| **Signed URL TTL**             | 5 minutes (300 seconds)       |
| **Job retry attempts**         | 3 (before terminal failure)   |
| **Export concurrency lock**    | 10 minutes                    |

## Required Roles

Minting operations require **admin** or **owner** role within the target brand workspace. Read-only operations (listing batches, tokens, events) are available to all authenticated members with brand access.

## Performance Benchmarks

The following benchmarks reflect observed throughput under real-world conditions:

| Scenario                          | Approximate Duration |
| --------------------------------- | -------------------- |
| 100,000 units (single job)        | \~5–6 seconds        |
| 1,000,000 units (single job)      | \~60–90 seconds      |
| 2,500,000 units (concurrent jobs) | \~3 minutes total    |

### Scaling Characteristics

* **Linear scaling**: throughput scales linearly with job size under tested concurrency.
* **No over-mint**: the system guarantees that `minted_quantity` never exceeds `requested_quantity`.
* **No duplicate identities**: serial numbers and token IDs are unique across all jobs.
* **Deterministic tracking**: `minted_quantity` on each batch accurately reflects the number of units generated.

## Pagination

All list endpoints use deterministic cursor-based pagination:

* Cursors are composite (timestamp + ID) for tie-breaking.
* Default page sizes vary by endpoint (typically 20–50).
* The `has_more` flag indicates whether additional pages exist.
* Cursor values for the next page are included in every response.

This ensures stable pagination even when new records are being created concurrently.

## Idempotency

* Mint submissions support client-provided idempotency keys.
* Duplicate submissions return the existing job without creating new records.
* Idempotency keys are unique per brand workspace.

## Notes

* If a production batch is linked and has a quantity set, the requested quantity must not exceed the production batch quantity.
* If no production batch is linked, only the platform limit (1,000,000 units) applies.
* In serial mode, lot entries must not be provided.
* In lot mode, the sum of all lot quantities must equal the total requested quantity.

## Related Docs

* [Minting Overview](/minting/overview)
* [Minting Architecture](/minting/architecture)
* [Minting Lifecycle](/minting/lifecycle)
* [Minting FAQ](/minting/faq)