generated: '2026-09-01'
method: searched
source: https://docs.0xarchive.io/rate-limits
sources:
- https://docs.0xarchive.io/rate-limits
- https://docs.0xarchive.io/errors
- https://docs.0xarchive.io/websocket/tier-limits
- openapi/0xarchive-openapi.json (components.responses.RateLimited)
limit_count: 6
note: >-
  0xArchive throttles on five separate axes rather than one - requests per second, concurrent
  queries, monthly credits, WebSocket subscription count and replay speed - plus a plan-scoped
  history window that behaves like a limit even though it is not a rate. All of them are
  per-account and published per tier. The single caveat worth flagging for an agent: the
  X-RateLimit-* headers below are DECLARED in the OpenAPI 429 response, but they were not
  observable on an unauthenticated request - a 401 and a 200 /health response on 2026-09-01
  carried x-request-id but no RateLimit family header, so an agent cannot read remaining budget
  until it is authenticated and, on this evidence, may only see it on a 429.
limits:
- scope: per-account
  control: requests per second
  window: 1 second
  limit:
    Free: 15
    Build: 50
    Pro: 150
    Scale: 500
    Enterprise: from 1,000
  burst: null
  note: Burst rate across REST calls. The docs prescribe backoff and queuing rather than a stated burst allowance.
- scope: per-account
  control: concurrent queries
  window: instantaneous
  limit:
    Free: 3
    Build: 10
    Pro: 20
    Scale: 20
    Enterprise: from 500
  note: Applies to long-running or heavy history calls; the docs advise bounding worker pools.
- scope: per-account
  control: monthly API credits
  window: calendar month
  limit:
    Free: 50000
    Build: 80000000
    Pro: 400000000
    Scale: 2000000000
    Enterprise: unlimited
  note: >-
    Credits are consumed by data volume and route cost, not by request count. L3/L4, replay and
    deep-history routes cost materially more than shallow market-state routes.
- scope: per-account
  control: WebSocket subscriptions
  window: concurrent
  limit:
    Free: 10
    Build: 500
    Pro: 3000
    Scale: 20000
    Enterprise: unlimited
- scope: per-account
  control: WebSocket replay speed multiplier
  window: per replay stream
  limit:
    Free: 10x
    Build: 50x
    Pro: 100x
    Scale: 300x
    Enterprise: from 500x
  note: >-
    Hyperliquid core l4_diffs and l4_orders replay ignores the speed parameter entirely - it uses
    a checkpoint-anchored bulk stream.
- scope: per-account
  control: history window / request span
  window: per request or per replay
  limit:
    Free: most recent rolling 30 days, maximum 30-day span per request or replay
    Build: full retained archive
    Pro: full retained archive
    Scale: full retained archive
    Enterprise: full retained archive
  note: >-
    The only route-shaped restriction in the plan model. Every tier including Free reaches every
    route family, schema and served depth; only the reachable time range differs.
exhaustion:
  status_code: 429
  error_code: rate_limited
  envelope: '{"success": false, "error": {"code": "rate_limited", "message": "Rate limit exceeded"}, "meta": {"request_id": "<uuid>"}}'
  note: >-
    429 is returned for rate, concurrency AND credit exhaustion - the status alone does not
    distinguish which of the three was hit.
response_headers:
- header: X-RateLimit-Limit
  type: integer
  meaning: Requests per second limit
  source: openapi/0xarchive-openapi.json components.responses.RateLimited.headers
  observed_anonymously: false
- header: X-RateLimit-Remaining
  type: integer
  meaning: Remaining requests this second
  source: openapi/0xarchive-openapi.json components.responses.RateLimited.headers
  observed_anonymously: false
- header: X-RateLimit-Reset
  type: integer
  meaning: Unix timestamp when the limit resets
  source: openapi/0xarchive-openapi.json components.responses.RateLimited.headers
  observed_anonymously: false
- header: Retry-After
  meaning: >-
    Documented as the authoritative retry signal on 429 - "Honor Retry-After when the response
    includes it" - but the docs are explicit that it is not always present, and it does not appear
    in the OpenAPI 429 response headers block.
  source: https://docs.0xarchive.io/errors
  observed_anonymously: false
  conditional: true
- header: x-request-id
  meaning: Per-request correlation UUID, present on every response including errors.
  source: probed live on https://api.0xarchive.io/health and an unauthenticated /v1/instruments call
  observed_anonymously: true
retry_policy:
  retryable:
  - 429
  - transient 5xx
  - network timeouts
  non_retryable_unchanged:
  - 400 malformed request or invalid parameter
  - 401 missing or invalid API key
  - 403 refused request
  - 404 unknown route or symbol
  strategy: >-
    Honor Retry-After when present; when absent use capped exponential backoff with jitter,
    reduce concurrency, and narrow the request window before widening the job again. Retry
    within a bounded budget only.
  source: https://docs.0xarchive.io/errors
probe:
  checked: '2026-09-01'
  evidence:
  - url: https://api.0xarchive.io/health
    status: 200
    ratelimit_headers_present: false
    headers_observed: [x-request-id, strict-transport-security, x-content-type-options, x-frame-options, access-control-expose-headers]
  - url: https://api.0xarchive.io/v1/instruments
    status: 401
    ratelimit_headers_present: false