We set our API rate limits in October 2024 based on what we thought were reasonable numbers. We were wrong. The limits were too low for real SaaS workloads and even for developers doing serious integration testing.

So we fixed it. As of January 2025, the new limits are:

Plan	API calls / month	Rate limit	Change
Sandbox	5,000	10/sec	Same
Dev	500,000	20/sec	25x increase
Prod	Unlimited	40/sec	No cap

We also extended the Sandbox evaluation window from 3 days back to 30 days of inactivity before deletion.

Why we made this change

Two things became clear after the October pricing update.

First, the limits were simply too low for real workloads. A SaaS company running transaction categorization across multiple end-customers can easily generate 10,000 predictions per day. At 20,000 calls per month on a Dev instance, you hit the wall in two days. That is not a development environment. That is a demo that expires. Prod at 200,000 had the same problem at scale. Customers processing 50,000 invoices per month need predictions, data uploads, and search queries. The limit forced conversations about call budgets instead of conversations about prediction quality.

Second, and more fundamentally: at this stage of the product, being generous is the right strategy. We are building a new category. The most important thing right now is that developers and SaaS teams can use Aito without friction, discover what a predictive database does for their use case, and build on it. Restrictive limits work against that. We would rather have customers pushing the boundaries of what they can build with Aito than carefully rationing API calls.

What the new limits mean in practice

Dev at 500,000 calls per month means you can run a realistic integration test with production-like volumes. Upload a dataset, run a few thousand predictions, iterate on your query structure, and still have capacity left. This is what a development instance should allow.

Prod at unlimited means the predictive database scales with your workload. No per-call cost anxiety. No throttling surprises at month-end. You pay a flat monthly rate and query as much as you need. For SaaS companies embedding Aito predictions into their product, this is the model that makes sense.

Sandbox at 30 days means you have time to evaluate properly. The 3-day inactivity window was too aggressive. If you upload data on Monday and get pulled into other work, your instance should still be there when you come back the following week.

What did not change

Pricing stays the same: Sandbox is free, Dev is EUR 75 per month, Prod is EUR 350 per month. Storage limits, active server hours, and all other plan features remain unchanged.

The rate limits per second (10, 20, 40) also stay the same. These protect instance stability. The change is entirely about monthly volume caps.

How are Aito's API calls counted?

Every request to the Aito API counts as one call. This includes data uploads, schema operations, predictions, recommendations, and searches. Requests through the Console UI or the Python SDK use the same API and count the same way.

The monthly quota resets at UTC midnight on your instance creation anniversary date. You can monitor your usage in the Console dashboard.

Current rate limits for each Aito plan

	Sandbox (free)	Dev (EUR 75/mo)	Prod (EUR 350/mo)
API calls / month	5,000	500,000	Unlimited
Rate limit / second	10	20	40
Storage	100 MB	500 MB	1 GB (extendable)
Active server / day	3 hours	12 hours	24/7
Inactivity deletion	30 days	N/A	N/A

If you are on a Dev or Prod plan, the new limits are already active. No action needed on your part.

If you have been holding off on evaluating Aito because the trial window was too short, create a free Sandbox. You have 30 days and 5,000 API calls to work with. That is enough to upload a real dataset, run predictions, and decide if a predictive database fits your use case.

Questions? Reach us at support@aito.ai.

Back to blog list