generated: '2026-08-12'
method: searched
source: https://docs.epsilon3.io/#rate-limiting
limit_count: 3
note: >-
  Epsilon3 publishes both the numeric limits AND the runtime response headers, which is
  the rarer and more useful half. Limits are stated for the REST API and, separately,
  for telemetry data_update events on the realtime SocketIO namespace.
limits:
- id: rest-api
  scope: per-api-key
  scope_detail: >-
    Applied per API key (team or user) for API-authenticated requests, and per user for
    web-authenticated requests.
  limit: 15000
  window: 10 minutes
  unit: requests
  burst: null
  applies_to: https://api.epsilon3.io/v1 (all authenticated endpoints)
  exhaustion_status: 429
  exhaustion_body: null
  source: https://docs.epsilon3.io/#rate-limiting
- id: telemetry-team
  scope: per-team
  limit: 1000
  window: 1 second
  unit: data_update events
  applies_to: SocketIO namespace /v1/telemetry/realtime, across all active streams
  exhaustion_status: null
  exhaustion_behavior: >-
    Excess data_update events are SILENTLY DROPPED - the value is not processed and no
    error is returned on the emitting call. The drop is reported out-of-band on the
    telemetry_connection_status event, which includes warnings naming the limit exceeded
    and how many events were dropped.
  source: https://docs.epsilon3.io/#rate-limits
- id: telemetry-stream
  scope: per-stream
  scope_detail: applied independently per stream_id
  limit: 4
  window: 1 second
  unit: data_update events
  applies_to: SocketIO namespace /v1/telemetry/realtime
  exhaustion_behavior: same silent drop plus telemetry_connection_status warning
  source: https://docs.epsilon3.io/#rate-limits
headers:
  published: true
  standard: draft-ietf-httpapi-ratelimit-headers (RateLimit-* form, unprefixed)
  response_headers:
  - name: RateLimit-Limit
    description: Maximum number of requests allowed in the current window
  - name: RateLimit-Remaining
    description: Number of requests remaining in the current window
  - name: RateLimit-Reset
    description: Time in seconds until the current rate limit window resets
  always_present: true
  always_present_note: Epsilon3 states these headers are included in EVERY response, not
    only on 429.
  retry_after: false
  retry_after_note: >-
    No Retry-After header is documented. An agent must compute its backoff from
    RateLimit-Reset.
payload_limits:
- id: telemetry-data-update-size
  limit: 100KB
  applies_to: each data_update payload on /v1/telemetry/realtime
  behavior: payloads exceeding the limit are rejected and not processed; split across
    multiple data_update events
  source: https://docs.epsilon3.io/#data-update-event
- id: command-timeout
  limit: 18000 seconds (5 hours)
  applies_to: maximum asynchronous command expiration
  default: 55 seconds when no correlation_id is supplied (synchronous)
  source: https://docs.epsilon3.io/#command-timeout
on_prem:
  note: >-
    Epsilon3 states on-premises deployments may have different rate limit thresholds
    configured by the customer's own administrator, so these figures describe the
    multi-tenant cloud only.