> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.anthropod.in/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.anthropod.in/_mcp/server.

# Errors and limits

> Handle failures consistently and keep reads bounded.

This page describes the data-read APIs. Existing batch, outbound calling, and embedded URL APIs retain the error formats documented in their own guides.

Errors use the HTTP status and the `application/problem+json` media type. Preserve `request_id` when contacting support.

```json
{
  "type": "about:blank",
  "title": "Bad Request",
  "status": 400,
  "detail": "conversation_time must use epoch seconds, not milliseconds.",
  "code": "INVALID_ARGUMENT",
  "request_id": "req_demo_001"
}
```

| Status    | Meaning                                               | Client action                                                 |
| --------- | ----------------------------------------------------- | ------------------------------------------------------------- |
| 400       | Invalid parameter, range, or cursor                   | Correct the request; restart the query for an invalid cursor. |
| 401       | Missing or invalid credentials                        | Check the assigned API key.                                   |
| 403       | Service or Voice Agent is outside your client account | Use a service or agent owned by your account.                 |
| 404       | Resource unavailable within the authorized scope      | Check the ID and service.                                     |
| 429       | Rate limit exceeded                                   | Honor `Retry-After`.                                          |
| 500 / 503 | Temporary server failure                              | Retry reads with bounded exponential backoff and jitter.      |

The default page size is 20, with a maximum of 50 records and a date window of up to 31 days. The defaults are 300 requests per minute per IP, 30 per standard endpoint and export retrieval per API key, and 5 for export creation per API key; see [rate limits](/rate-limits). Cursors last 15 minutes. Export downloads remain available for seven days.