# Rate limits

Source: https://developers.novilabs.com/platform/rate-limits/

> Request rates, concurrency, size limits and back-off rules for each Novi product, as published today.

Only the EVA REST API publishes a request rate today; the others limit work by size, jobs in flight or warehouse.

## At a glance

| Product | Request rate | Work in flight | Size limits | Rate headers |
| --- | --- | --- | --- | --- |
| Novi UDM Data (v4) | Not published yet | Not published yet | Not published yet | Not published yet |
| Novi Forecasts API | Not published yet | Async forecasts: 50 per company. Plan economics: 4 per company. Exports: 20 per model set, 3 per plan. Your unfinished plan forecasts: `limits.plan_unfinished_forecasts` | 50 drawn wells per `POST /forecasts`. Bodies: 64 MB on `POST /plans`, 10 MB on `POST /exports` | `Retry-After` only, when a job cap is reached |
| Novi EVA REST API | 10 requests per second on average, bursts up to 30 | 5 requests | 50,000 wells per `POST /v1/data`, 1,000 per `POST /v1/production` | Not published yet |
| Novi EVA MCP server | Not published yet | A few data tool calls at a time per person; more wait briefly | Request body 1 MiB. Data tool result 64 KiB. Chart result 256 KiB. 30-second deadline per data tool call | Not published yet |
| Novi M&A Intelligence MCP server | `ma_search_deals` is rate limited per person; the value is not published yet | Not published yet | 100 rows per page; 50 deals in your basket | Not published yet |

## Novi UDM Data (v4)

Your queries run on a Snowflake warehouse, so compute, not a request count, sets the pace.

- In a reader account, the first query after about a minute idle takes 10 to 30 seconds to start. The same SQL run again within 24 hours returns at once.
- On your own Snowflake account, your warehouse runs the queries and you pay for it.
- On a BI or lakehouse connection, statements stop after 2 hours.

## Novi Forecasts API

`GET /dataset` `limits` and `GET /export-formats` `limits` report most limits: read them rather than hard-coding values. Every limit and how to back off: [Errors and limits](https://developers.novilabs.com/forecasts/guides/errors-and-limits/).

## Novi EVA

- **REST:** stay under 10 requests per second and 5 in flight. Batch wells into one call instead of many small ones.
- **MCP:** `LIMIT_EXCEEDED` and `DEADLINE_EXCEEDED` are retryable. Wait, then retry, or narrow the request.

## Planned

| Item | Status |
| --- | --- |
| `RateLimit` and `RateLimit-Policy` headers, so a client can pace itself | [Planned] |
| Published request rates and concurrency for Novi UDM Data over HTTPS | [Planned] |
