generated: '2026-08-19'
method: searched
source: https://docs.abloatai.com/pricing + https://docs.abloatai.com/errors + https://docs.abloatai.com/mcp
note: >-
  Ablo signals two different things on two different header families, and they are not the same
  budget. The X-Usage-* family is the BILLING meter (monthly operations against the tier floor) and
  is documented on the pricing page. The x-ratelimit-* family is a per-IP request throttle, observed
  live on the public MCP endpoint. A per-key request-rate limit is documented only by its error
  (rate_limit_exceeded, with Retry-After) — the numeric value is not published.
limit_count: 6
limits:
  - name: monthly operations (Free)
    scope: per-organization
    window: monthly
    limit: 100000
    burst: null
    enforcement: hard stop — requests past it are refused rather than billed
    tier: Free
    source: https://docs.abloatai.com/pricing
  - name: daily operations (Free)
    scope: per-organization
    window: daily
    limit: 10000
    enforcement: burst guard — spreads the monthly allowance over at least 10 days
    tier: Free
    source: https://docs.abloatai.com/pricing
  - name: monthly operations (Scale)
    scope: per-organization
    window: monthly
    limit: 10000000
    enforcement: metered above the floor, not stopped
    tier: Scale
    source: https://docs.abloatai.com/pricing
  - name: concurrent WebSocket connections
    scope: per-principal or per-organization
    window: concurrent
    limit: 25
    tier: Free
    enforcement: 'connection_limit_exceeded (429, retryable)'
    source: https://docs.abloatai.com/pricing
  - name: concurrent WebSocket connections
    scope: per-principal or per-organization
    window: concurrent
    limit: 1000
    tier: Scale
    enforcement: 'connection_limit_exceeded (429, retryable)'
    source: https://docs.abloatai.com/pricing
  - name: integration-helper MCP requests
    scope: per-IP
    window: 1m
    limit: 120
    burst: null
    endpoint: https://www.abloatai.com/api/mcp
    method: probed
    enforcement: 'Rate-limit headers echoed on every response.'
    source: https://docs.abloatai.com/mcp
undocumented:
  - name: per-key request rate limit
    scope: per-api-key
    limit: null
    note: >-
      The error registry documents rate_limit_exceeded — "This API key is sending requests faster
      than its rate limit allows. Slow down and retry after the delay in the Retry-After header" —
      so a per-key throttle is enforced, but no numeric limit or window is published anywhere in the
      docs. An agent cannot budget against it in advance; it can only react to the 429.
response_headers:
  billing_meter:
    scope: 'Documented as present on every metered response.'
    method: searched
    headers:
      - {header: X-Usage-Limit, meaning: operations included in the period, example: '50000000'}
      - {header: X-Usage-Used, meaning: operations consumed, example: '12480311'}
      - {header: X-Usage-Remaining, meaning: operations left, example: '37519689'}
      - {header: X-Usage-Reset, meaning: ISO 8601 timestamp the period rolls over, example: '2026-08-01T00:00:00.000Z'}
  request_throttle:
    scope: https://www.abloatai.com/api/mcp
    method: probed
    headers:
      - {header: x-ratelimit-limit, observed_value: '120'}
      - {header: x-ratelimit-remaining, observed_value: '118'}
      - {header: x-ratelimit-reset, observed_value: '1787141193', format: unix epoch seconds}
  retry:
    method: searched
    headers:
      - {header: Retry-After, emitted_on: 'rate_limit_exceeded (429)'}
exhaustion:
  status_code: 429
  codes:
    - {code: rate_limit_exceeded, http_status: 429, retryable: true, remedy: 'Slow down and retry after the delay in the Retry-After header.'}
    - {code: quota_exceeded, http_status: 429, retryable: true, remedy: 'Succeeds again once the quota resets or the limit is raised.'}
    - {code: connection_limit_exceeded, http_status: 429, retryable: true, remedy: 'Close idle connections, or retry once others drain.'}
  spec_evidence: 'Every one of the 32 operations in the OpenAPI declares a 429 response returning ErrorEnvelope.'
metering_exemptions:
  - 'Only successful requests are metered; failures are free.'
  - 'A retry carrying the same Idempotency-Key is metered once.'
  - 'Reads served from the client''s local pool never reach the engine and never count.'
x-evidence:
  - {url: 'https://docs.abloatai.com/pricing', http_status: 200}
  - {url: 'https://www.abloatai.com/api/mcp', http_status: 200, note: 'POST tools/list returned x-ratelimit-limit: 120'}