Rate limits

Limits are per API token, so one busy integration doesn't slow down another.

	Limit
Requests	120 a minute per token
File uploads (`POST …/media`)	30 a minute per token, within the 120
Results per page	25 by default, at most 100 (`page[size]`)
Files in a post	20
Accounts in a post	50

Every response says where you stand:

X-RateLimit-Limit: 120
X-RateLimit-Remaining: 117

Past the limit, requests answer 429 with a Retry-After header (seconds) and a JSON:API error:

{
  "errors": [
    {
      "status": "429",
      "code": "too_many_requests",
      "title": "Too many requests",
      "detail": "Slow down: see the Retry-After header for when to try again."
    }
  ]
}

Wait that many seconds before trying again. For bursts, such as importing many posts, pace requests rather than retrying in a loop.

MCP

AI apps connected through the MCP server have their own limit: 120 tool calls a minute per person, across all of that person's apps.

Webhooks

Refract waits up to 10 seconds for your endpoint to answer each delivery and retries failed ones on its own schedule (see Webhook events), so receivers should answer quickly and do slow work afterwards.