# How to Batch CS2 Price API Requests

> For bulk current prices, download `GET /v1/prices/latest` once and partition its `items` map locally. For historical endpoints, submit exact names in chunks of at most 100.

The current snapshot already contains every tracked item and all six marketplaces, so many selected-item requests would produce different response times without improving coverage. One snapshot gives the batch a common `response_time`.

## Data and limits for bulk processing

- `GET /v1/schema` supplies the canonical `market_hash_name` keys accepted across price, history, archive, and liquidity data.
- `GET /v1/prices/latest` returns the complete current-price snapshot in one large response. It requires `Accept-Encoding: gzip`.
- POST history and archive endpoints accept up to 100 items and request bodies up to 1 MiB.
- All plans include unlimited requests with a rate limit of 10 requests per second per user.
- POST endpoints can return successful `items` beside per-item `errors[]`, so partial success must remain attached to each chunk.

## How to organize a batch

For current prices, fetch the full snapshot, validate `currency` and `items`, and select the required names from the returned map. Keep the snapshot's `response_time` and each source's `collected_at`; they answer different freshness questions.

For a history or archive job, deduplicate exact names, split them into chunks of no more than 100, and retain every chunk's normalized `start`, `end`, and `interval`. Merge results by exact item key and keep `errors[]` instead of converting unresolved names into zero prices.

Use bounded retries for request-wide `429` and `5xx` failures. An item-level `unknown_item` or `invalid_format` needs a corrected schema name, not an unchanged retry.

## Applicable endpoints

| Endpoint | Returns | Plans |
| --- | --- | --- |
| `GET /v1/prices/latest` | One current all-item, all-source snapshot | All plans |
| `GET /v1/schema` | Exact item names and catalog metadata | All plans |

## Batch boundaries

- Separate current snapshots fetched at different times are not an atomic market view. Prefer one download when comparisons need a common response time.
- A missing or null source field is missing data. Never replace it with zero during a merge.
- The full current snapshot is large. Stream, cache, or filter it after gzip decompression when memory is constrained.
- Endpoint-specific history starts and interval limits still apply to every chunk; dividing a request cannot extend coverage.

See [the current snapshot cache guide](/resources/how-to-download-and-cache-all-current-cs2-prices) for repeated bulk reads.