generated: '2026-09-12'
method: probed
source: >-
  openapi/aeternity-node-openapi.yml (components.responses.ServiceUnavailable),
  https://github.com/aeternity/ae_mdw/blob/master/README.md,
  live header inspection of https://mainnet.aeternity.io/v3/status and https://mainnet.aeternity.io/mdw/v3/status (2026-09-12)
summary: >-
  No numeric request quota is published for either public gateway, and no rate-limit headers are returned.
  What the provider does publish is a backpressure contract — a 503 with Retry-After declared on every node
  operation — and a set of documented WebSocket connection ceilings for the middleware.
response_headers:
  rate_limit_headers: []
  observed:
  - x-ae-height
  - x-request-id
  detail: >-
    Live responses from both gateways carry no X-RateLimit-*, RateLimit-* or Retry-After header at rest. The
    Retry-After header appears only on the 503 overload response.
exhaustion:
  status: 503
  header: Retry-After
  schema: integer seconds, minimum 0, example 3
  applies_to: all 79 node operations
  quote: >-
    "The node is temporarily unable to handle the request: either its request queues are saturated (overload
    protection) or it has not yet reached a stable mode. Overload protection is applied to every endpoint, so
    this response can be returned by any operation. The Retry-After header tells the client how long to back
    off before retrying."
  source: openapi/aeternity-node-openapi.yml components.responses.ServiceUnavailable
rate_limits:
- scope: per-connection (middleware WebSocket)
  limit: 100000
  unit: subscriptions per connection
  setting: MAX_SUBS_PER_CONN
  note: Documented default in the ae_mdw README configuration table; operator-configurable, and the value in force on the public gateway is not separately published.
  source: https://github.com/aeternity/ae_mdw/blob/master/README.md
- scope: per-deployment (middleware WebSocket)
  limit: 1000
  unit: concurrent WebSocket connections
  setting: MAX_WS_CONNECTIONS
  source: https://github.com/aeternity/ae_mdw/blob/master/README.md
- scope: per-ip (middleware WebSocket)
  limit: 50
  unit: concurrent WebSocket connections from one IP
  setting: MAX_WS_CONNECTIONS_PER_IP
  behaviour: >-
    "Connections may now be rejected at the WebSocket handshake stage if server-configured limits are
    exceeded (total connections, connections per IP). The server closes the socket with a normal closure in
    this case. Clients should handle CLOSE frames and apply exponential back-off before reconnecting."
  source: https://github.com/aeternity/ae_mdw/blob/master/README.md
- scope: per-deployment (middleware WebSocket)
  limit: 2000000
  unit: total subscriptions across all connections
  setting: MAX_TOTAL_WS_SUBS
  source: https://github.com/aeternity/ae_mdw/blob/master/README.md
- scope: per-connection (middleware WebSocket)
  limit: 2000
  unit: queued messages per client before shedding
  setting: MAX_WS_CLIENT_BACKLOG
  source: https://github.com/aeternity/ae_mdw/blob/master/README.md
- scope: per-request (middleware GraphQL)
  limit: 1000
  unit: Absinthe query complexity score; queries exceeding it are rejected
  setting: GRAPHQL_MAX_COMPLEXITY
  source: https://github.com/aeternity/ae_mdw/blob/master/docs/graphql.md
- scope: per-request (node dry-run)
  limit: null
  unit: total gas per dry-run request, capped by a node-configured limit
  behaviour: Exceeding it returns 403 "Over the gas limit".
  note: The cap is node-configured and its value on the public gateway is not published.
  source: openapi/aeternity-node-openapi.yml ProtectedDryRunTxs
limit_count: 7
http_limit_count: 0
note: >-
  http_limit_count is 0 on purpose: æternity publishes no request-per-second/per-hour HTTP quota for
  mainnet.aeternity.io, no terms of use for the public gateway, and no header from which one could be read.
  Every limit recorded above is a WebSocket/GraphQL/gas ceiling documented as a configuration default in the
  provider's own repository, not a quota the gateway advertises at runtime. An agent should treat 503 +
  Retry-After as the only authoritative throttling signal.