> ## Documentation Index
> Fetch the complete documentation index at: https://docs.sharedgraph.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Anonymous limits

> Per-address budgets and how to handle throttling.

Anonymous reads are a pilot convenience for public content. They may later require an application credential. Application-only read credentials are a planned follow-up, not a currently supported grant.

| Budget | Limit | Applies to |
| - | - | - |
| Combined anonymous reads | 120 requests per minute per connecting address | All supported anonymous social GET requests |
| Shared paging budget | 30 requests per minute per connecting address | `/api/events` and `/api/reconcile` together |

Paging requests consume both budgets. Requests behind the same connecting address share limits. The network uses the edge-supplied connecting address and does not store the raw address in the limiter.

## Handle HTTP 429

Exhausting either budget returns HTTP 429 with error `pilot_rate_limited` and a `Retry-After` header containing positive integer seconds until the window resets. Wait at least that long, then retry with bounded backoff. Avoid concurrent retry loops that consume the shared budget immediately.

If limit storage is unavailable, admission fails closed rather than allowing unlimited reads. Treat HTTP 503 as temporary unavailability and retry with bounded backoff.

## Authorized requests

Authorized reads are exempt from these anonymous budgets. Existing per-actor and per-application write quotas still apply. Sending an invalid or expired token does not convert a request into anonymous access. Never fall back to anonymous reads inside a signed-in experience after a refusal.

These limits regulate service access; they do not make public posts confidential. Network availability, deletion and moderation rules continue to apply.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.