Skip to main content

Limits

The Mastra Memory Gateway enforces rate limits and pagination constraints to protect the service and ensure fair usage.

Rate limits​

Rate limits are applied per API key. When a limit is exceeded, the gateway returns a 429 status code with a Retry-After header.

LLM proxy​

Applies to POST /v1/chat/completions, POST /v1/messages, and POST /v1/responses.

RuleDefault limitWindow
Global (all endpoints)5,000 requests60 seconds
LLM proxy2,000 requests60 seconds
LLM burst400 requests10 seconds

Memory API​

Applies to the /v1/memory/* endpoints. Read and write operations have separate limits.

RuleDefault limitWindow
Memory read (GET, HEAD)1,200 requests60 seconds
Memory write (POST, PATCH, DELETE)600 requests60 seconds

Rate limit headers​

Every response includes rate limit headers:

HeaderDescription
X-RateLimit-LimitMaximum requests allowed in the current window
X-RateLimit-RemainingRemaining requests in the current window
X-RateLimit-ResetUnix timestamp when the window resets
X-RateLimit-ScopeWhich limit applies (llm_proxy, memory_read, or memory_write)
Retry-AfterSeconds until the limit resets (only present on 429 responses)

Rate limit error response​

{
	"error": {
		"message": "Rate limit exceeded. Please retry after the reset time.",
		"type": "rate_limit_error",
		"scope": "llm_proxy",
		"retry_after_seconds": 12
	}
}

Pagination limits​

List endpoints accept limit and offset query parameters.

ParameterDefaultMaximum
limit (threads, messages)50200
limit (observation history)10200
offset0—

Authentication​

  • All API keys use the msk_ prefix
  • An invalid or missing API key returns a 401 error

Error format​

All error responses follow the same structure:

{
	"error": {
		"message": "Human-readable error description",
		"type": "error_type"
	}
}

Common error types:

TypeStatus codeDescription
authentication_error401Invalid or missing API key
not_found404Thread or resource not found
rate_limit_error429Rate limit exceeded
invalid_request_error400Malformed request or unsupported provider/endpoint combination