Docs
Agents and API

Request limits

Request rates per plan, and how to handle a 429 RATE_LIMITED response.

Xentropy limits how many requests each account can make per minute. Request limits are separate from plan quotas (how many artifacts, dashboards or exports you can keep or create) and from fair-use ceilings. Those return QUOTA_EXCEEDED or FAIR_USE_LIMIT, not RATE_LIMITED.

Rates per plan

Each limit counts requests in a fixed one-minute window, per account.

LimitWhat it countsFreePro
API requestsEvery authenticated REST request (session or API key)600 / min1,200 / min
Expensive data requestsData and dataset previews, schema reads, plan previews, OpenAPI document fetches, connection tests, S3 object listings and file uploads30 / min120 / min
MCP requestsEvery request to the MCP endpoint300 / min600 / min

An expensive data request also counts as an API request.

Some limits apply per network address rather than per account:

LimitRate
Failed API keys or access tokens from one address30 / min. After that, every request from that address is refused until the minute ends, including requests with a valid key.
Unauthenticated endpoints, such as the plan catalog and connection sign-in callbacks120 / min
Shared and embedded charts and dashboards120 page views / min and 600 file downloads / min

Sign-in, magic links, profile changes and feedback have their own smaller limits to protect your account.

Folder limits

Folders have size limits, not request rates: 1,000 folders per account, 5 levels of nesting and 100 items per move. They are the same on every plan. Going over returns FOLDER_LIMIT_REACHED, FOLDER_DEPTH_EXCEEDED or, for a move of more than 100 items, a validation error. See Folders.

The 429 response

Every request limit answers the same way: HTTP 429 with the code RATE_LIMITED.

HTTP/1.1 429 Too Many Requests
Retry-After: 17
RateLimit-Policy: "api.expensive";q=30;w=60
RateLimit: "api.expensive";r=0;t=17
Content-Type: application/json

{
  "error": {
    "code": "RATE_LIMITED",
    "message": "Too many requests. Retry after the Retry-After interval.",
    "policy": "api.expensive",
    "retryAfter": 17,
    "retryable": true
  },
  "requestId": "…"
}
  • Retry-After and error.retryAfter give the seconds until the window resets.
  • RateLimit-Policy names the limit (q requests per w seconds), following the IETF RateLimit header fields draft.
  • RateLimit shows r=0 requests remaining and t seconds until the reset.
  • error.policy names the limit that applied: api.user, api.expensive, mcp.user, auth.bad-credential, public.anon, public.page or public.file.

Over MCP, the same fields appear in the JSON error body of the HTTP 429 response, with the same headers.

Handling a 429

  1. Wait at least Retry-After seconds before retrying. Retrying sooner only returns another 429.
  2. Add jitter when several workers share one API key, so they don't all retry at the same moment.
  3. Spread expensive data requests out. Reuse a preview or a dataset version instead of requesting it again.
  4. If you hit auth.bad-credential, check the key: it was revoked, mistyped or expired. Create a new one in API keys.

If your workload needs higher rates, Pro doubles the API rate and quadruples the expensive data rate.

On this page