Xata applies some limits based on the plan level and the API endpoint in order to ensure reliable operation and scalability.

API keys per user40
Workspaces per user40
Databases per workspace40
Tables per branch100
Tables per workspace1000
Database name length255 chars
Branch name length255 chars
Number of columns per table256
id column length255 chars
String column length2048 chars
Text column size200 KB
Multiple type total size65536 chars
Record size600 KB
Request size20 MB
Bulk insert1000 records
Transaction1000 operations
Page size1000 records
Offset pagination size50000 records
Cursor pagination depthUnrestricted
File size (when using file binary APIs)1 GB
File size (when using Record APIs)20 MB
Query response size (when it includes file content)20 MB
Number of files in a file array50
Max ttl for a signed URL24h

The REST API may return HTTP response code 429 as a throttling error when the rate limiting conditions are met:

  • Request rate limit: Total number of atomic HTTP requests submitted per second to any API endpoint under a branch
  • Concurrency limit: Number of requests executing in parallel at any given moment within a certain Store of a branch

For more information on rate limits, see our docs.

Xata uses store types with individual concurrency limits:

Store typeConcurrent requestsEndpoints
Consistent transactional store2data, bulk, transaction, query, summarize
Read replicas store3query, summarize with option consistency: eventual
Search & Analytics store5aggregate, search, ask, vectorSearch
File attachments store2data, file

In the event the concurrency limit in a store is exceeded, requests to it are throttled internally with a timeout of 100ms. After reaching the timeout, the API errors with HTTP response code 429.

For file attachments, the concurrency limits apply to CRUD and query operations on data and file endpoints. For content sharing and high throughput scenarios, the direct access URLs should be used, which are NOT subject to rate and concurrency limits.

Xata's SDK clients throttle and retry 429 requests until they can be processed. Instead of throwing an application error when you have a significant amount of traffic, it degrades the speed without failing. Whenever a request needs to be throttled by the SDK, it prints an application log with the amount of extra time it took to be processed. To avoid noticeable speed impact in your application, you can upgrade your plan.

Tips & best practices useful to optimize requests for the Xata API:

  • In case the client receives a 429 response code from Xata, it means the request was not processed so your application can safely retry it.
  • Log and track errors in your application error channels or log streams in order to monitor for throttling errors, so you can make informed decisions whether to upgrade your plan.
  • For writing records to Xata, use the bulk endpoint instead of the atomic (individual record insert) data endpoint. The bulk endpoint allows batching of multiple records in a single request, which favors the optimal use of both your branch's request rate limit and concurrency slots in the transactional store.
  • For requests to the query and summarize endpoints, consider using the consistency: eventual option to use the Read Replicas Store. This makes it less likely to reach the concurrency limit of either Store type and gets the best possible performance from your Xata plan.

Custom plans can increase or remove some limits. To discuss upgrading, contact us or complete the Contact Form.