50,000 requests/day rate limit scope

Imagery, data, and integrations
1

For tiles.mapillary.com, is the documented 50,000 requests/day rate limit scoped per application/client token, per Mapillary account, per IP address, or globally? If it is per application, can the quota be increased for a commercial production application?

Thanks for your support.

2

Hi @andynewmanny, thanks for the comment.

The different systems can block you at different levels, by IP and by app ID. The mentioned 50k is by app ID. At the same time, if you use the same IP address and this is a sudden spike, the system might block you earlier. Usually, this is more than enough for the average use case.

For other use cases, I would suggest contacting our support team at support@mapillary.com.

__
BR, Yaro

3

Hi @Yaro and @andynewmanny,

Thanks for this thread. We are running into a related issue here so I thought I’d share some data and ask a few questions.

We are a university research group (Makeability Lab, University of Washington) measuring how street-level imagery coverage changes over time. For this research, we sample a grid and a road network per city and read the z14 vector-tile census to count 360° panos. It runs nightly over a few cities from a single host.

Since 2026-08-12 we have been blocked three times on tiles.mapillary.com. Each time, it’s an HTTP 302 to the login page for every tile, which matches the by-IP layer described above. Note, we are well below the 50k/day by-app-ID limit (our heaviest day ever was ~10,700, about 21% of it).

Interestingly, the blocks do not seem correlated, necessarily, to overall daily volume. For example:

  • 2026-08-12 ~10,700 requests → blocked
  • 2026-08-14 ~26,400 requests → fine
  • 2026-08-20 ~5,000 requests → blocked
  • 2026-08-28 ~1,900 requests → blocked

The last block came while we were pacing at 60 requests/minute and capping the day at 3,500, which is roughly a seventh of the daily volume that had been fine two weeks earlier. So, perhaps there is a rolling window we are hitting or maybe Meta/Mapillary is more sensitive to ips that they’ve previously blocked?

Two questions:

1. Is there any guidance on what the per-IP layer actually keys on? Is it throughput of requests (i.e., num of requests/min) vs. overall daily volume? We are happy to restrict our requests to fit requirements.

2. For a coverage census like ours, is the vector-tile endpoint the right tool at all? If there is a bulk, export, or research data path better suited to “how many panos are in this bbox”, we would rather use that than page through tiles.

Thanks all,

Jon

1 Like