Monthly quotas
- API requests. Your plan sets a monthly API request quota, shown in the Editor under Settings > Usage. Cloud anchor operations (except attaching and removing content), geospatial anchor operations, scenes, projects and scene API requests each count as one request. When the quota is used up they answer 403
API_REQUEST_LIMIT_REACHEDuntil the team's usage period resets (its plan's renewal date, or the 1st for a team without a plan). File uploads, AI generation, co-location rooms, Advanced VPS and user tokens do not count. - Storage. Uploaded images, videos and 3D models, and AI outputs, count against the team's storage quota (403
STORAGE_LIMIT_REACHED). Anchor files (anchor-featuresandmeshuploads) do not. - AI credits. Each AI generation costs credits from the team's monthly AI credit budget; the AI operations list the costs (403
LIMIT_REACHED). - Compute time. Advanced VPS processing ($0.60 a minute, only for scans whose map is accepted), environment renders ($0.60 a minute) and co-location ($0.01 per device minute) are billed from one monthly compute pool. Plans include monthly credit ($5 on Basic, $15 on Pro), and usage beyond it is invoiced on the renewal date. Owners and admins set the monthly spend limit on the Usage page in the Editor (default $100); past it, billed operations answer 403
COMPUTE_SPEND_LIMIT_REACHED. See Plans, usage and compute billing. - Co-location needs a paid plan (403
PAID_PLAN_REQUIRED), and a project holds at most 500 rooms. The time devices spend in a relay room is billed as compute time, per device minute, against the team's monthly compute spend limit. The room operations themselves are not billed. - Advanced VPS. Creating a location or a scan and uploading a recording need a paid plan (403
PAID_PLAN_REQUIRED). Processing a scan is billed as compute time, only when its map is accepted, against the team's monthly compute spend limit; reads, downloads and image searches are not billed. Kept recordings and every file a scan builds count against the storage quota. An app user may start 10 scans a day unless the team set its own limit in the Editor (429VPS_USER_LIMIT_REACHED), and may have at most 3 scans waiting to be uploaded or processed (429VPS_USER_OPEN_SCANS_LIMIT).
Search bounds
| Operation | Client key | Secret and legacy keys |
|---|---|---|
| Nearby anchor searches (cloud and geospatial) | At most 50 results within at most 1,000 m. Cloud anchor searches return public anchors only. | At most 100 results within at most 5,000 m, once enforced. |
| Anchor lists (cloud and geospatial) | Not available. | At most 100 per page, once enforced. |
| File list | Not available. | At most 100 per page (secret keys only; legacy keys are refused). |
The default radius is 500 m and the default limit is 50. A larger limit or radius is reduced to the bound without an error; nearby searches echo the radius used in search.radius_meters, and lists echo the page size in pagination.limit.
Rate limits
| Requests | Limit | Counted per | Keys |
|---|---|---|---|
| Creating a cloud or geospatial anchor | 30 a minute | IP address and team | All |
Uploading an anchor file (anchor-features or mesh) | 30 a minute | IP address and team | All |
| Fetching one cloud or geospatial anchor | 300 a minute | IP address and team | All |
| Reporting a cloud anchor resolve | 300 a minute | IP address and team | All |
| Each other anchor operation (nearby, list, update, delete, attach, remove) | 600 a minute | IP address | All |
| Running a scene API request | 300 a minute | IP address | All |
| Running a scene API request | 120 a minute | Key | Secret |
| Running a scene API request | 120 a minute | Team | All callers; already enforced |
| All operations together | 1,200 a minute | Key | Secret |
| Co-location rooms | 600 a minute | Minting key and app user | User tokens |
| Minting user tokens | 1,200 a minute | Key | Secret; already enforced |
| Minting user tokens | 60 an hour | Key and app user | Secret; already enforced |
A request over a limit answers 429 RATE_LIMITED with a Retry-After header, in seconds.
The host and resolve limits are ARCore's Cloud Anchor limits, counted per IP address and team as ARCore counts them per IP address and Google Cloud project, so one team's busy venue never uses up another team's.
Per-key limits apply to secret keys only. Every copy of an app sends the same client key, so a per-key limit on it would let anyone who copied the key out of the app block the app for all its users. A client key is bounded by the per-IP limits and the team's quotas instead.
The limits not marked as already enforced, and the bounds marked "once enforced" above, start being enforced on a date announced in the changelog. Until then a request over them is served as usual.