generated: '2026-08-17'
method: searched
source: https://docs.api.memo.bank/topic/topic-rate-limiting
docs: https://docs.api.memo.bank/topic/topic-rate-limiting
limit_count: 0
note: >-
  Memo Bank documents the rate-limiting MECHANISM in full but publishes no NUMBERS. There is a dedicated
  rate-limiting topic that names the exhaustion status and all three response headers, yet states no
  requests-per-window figure, no burst allowance and no per-plan differentiation - so limit_count is 0 while
  the runtime signal an agent actually needs is fully specified. This is the better half of the trade to be
  on: a client that reads RateLimit-Remaining does not need the documented number, whereas a client given
  only a number and no headers cannot adapt. Recorded honestly rather than inferred.
signaling:
  documented: true
  headers_on_every_response: true
  headers:
  - name: RateLimit-Limit
    meaning: Total number of available requests between two quota resets.
  - name: RateLimit-Remaining
    meaning: Number of available requests until the quota is reset.
  - name: RateLimit-Reset
    meaning: Time remaining, in seconds, until the quota is reset.
  header_style: >-
    IETF draft RateLimit-* naming without the X- prefix, matching draft-ietf-httpapi-ratelimit-headers.
  retry_after: false
  retry_after_note: >-
    No Retry-After header is documented; clients are expected to use RateLimit-Reset instead.
exhaustion:
  status: 429
  reason_phrase: Too Many Requests
  body: Standard Memo Bank error envelope ({code, message}) - see errors/memo-bank-problem-types.yml.
  retry_guidance: >-
    Memo Bank lists 429 among the responses on which a caller should retry using the same Idempotency-Key,
    so a rate-limited mutating request is safe to re-send once the quota resets.
limits: []
limits_note: >-
  No published limit values. Memo Bank states only that "we enforce a rate limit on the number of HTTP
  requests that can be made in a given period" without quantifying the period or the count.
scope: not published
scope_note: >-
  The documentation does not say whether the quota is per application, per certificate, per workspace or per
  endpoint. Since credentials are issued per application (certificate + secret + private key) and IP
  allow-lists are configured per application, per-application is the most likely granularity, but Memo Bank
  does not state it and it is not recorded as fact here.
verification:
  probed: false
  reason: >-
    The API rejects all unauthenticated requests, and credentials are only issued by a Memo Bank banker to
    existing customers, so the headers could not be observed on a live 2xx response without an account.
    Recorded from documentation only.
gaps:
- No numeric limit, window or burst figure published anywhere in the API documentation.
- No statement of the rate-limit scope (per application / per workspace / per endpoint).
- No Retry-After header.
- Rate-limit headers are not described in the OpenAPI, only in the prose topic.
cross_links:
  conventions: conventions/memo-bank-conventions.yml
  errors: errors/memo-bank-problem-types.yml
  plans: plans/memo-bank-plans-pricing.yml