specification: API Commons Rate Limits
specificationVersion: '0.1'
schema: https://raw.githubusercontent.com/api-evangelist/interface-research/main/schema/api-commons.yml#/$defs/RateLimits
provider: Instantly
providerId: instantly-ai
generated: '2026-08-13'
method: searched
source: https://developer.instantly.ai/getting-started/rate-limit
also:
- https://github.com/Instantly-ai/instantly-starter-kit/blob/main/docs/conventions.md
- openapi/instantly-ai-api-v2-openapi.yml
created: '2026-05-22'
modified: '2026-08-13'
tags:
- Cold Email
- Outbound
- Sales
- Deliverability
- Rate Limiting
- Quotas
- Throttling
description: >-
  Instantly's published API rate limits. The global limits are workspace-wide — they are shared
  between API v1 and API v2 and across every API key in the workspace, so adding keys does not add
  headroom. A handful of endpoints carry their own, much lower limits. This file replaces the
  scaffold defaults written on 2026-05-22, which were invented placeholders (10/min free,
  120/min pro) and did not match anything Instantly publishes.
limit_count: 7
headers:
  limit: null
  remaining: null
  reset: null
  retryAfter: Retry-After
  policy: null
  note: >-
    Instantly documents NO X-RateLimit-* / RateLimit-* response headers. An agent cannot read its
    remaining budget from a response — the only runtime signal is the 429 status itself, plus a
    Retry-After header (in seconds) on the Google/Microsoft OAuth init endpoints. This is the
    weakest part of the API's runtime contract and the concrete thing for Instantly to fix.
responseCodes:
  throttled: 429
  quotaExceeded: 429
  upstreamThrottled: 503
limits:
- name: Global requests per second
  scope: workspace
  metric: requests_per_second
  limit: 100
  timeFrame: second
  applies:
  - https://api.instantly.ai/api/v2
  description: >-
    Workspace-wide. Shared between API v1 and API v2 and across all API keys in the workspace.
    Breaching either the per-second or the per-minute limit returns 429.
- name: Global requests per minute
  scope: workspace
  metric: requests_per_minute
  limit: 6000
  timeFrame: minute
  applies:
  - https://api.instantly.ai/api/v2
  description: Workspace-wide, shared across API v1 + v2 and all keys.
- name: List emails
  scope: endpoint
  operationId: listEmail
  endpoint: GET /api/v2/emails
  metric: requests_per_minute
  limit: 20
  timeFrame: minute
  description: Documented in the endpoint description — lower than the global limit.
- name: Send a test email
  scope: workspace-endpoint
  operationId: sendTestEmail
  endpoint: POST /api/v2/emails/test
  metric: requests_per_minute
  limit: 10
  timeFrame: minute
  description: 10 requests per minute per workspace. Can also return 200 with an error in the body.
- name: AI reply label test
  scope: workspace-endpoint
  operationId: testAiReplyLabelLeadLabels
  metric: requests_per_30_days
  limit: 500
  timeFrame: 30 days
  description: Test endpoint only; 500 requests per 30 days per workspace.
- name: OAuth sender connect init
  scope: workspace-endpoint
  operationId: initGoogleOAuth / initMicrosoftOAuth
  endpoint: POST /api/v2/oauth/{google|microsoft}/init
  metric: requests_per_minute
  limit: 75
  burst: null
  timeFrame: minute
  secondary_limit:
    scope: ip
    metric: requests_per_minute
    limit: 150
  description: >-
    75 per minute per workspace AND 150 per minute per IP, to comply with upstream provider limits.
    Returns 429 with a Retry-After header in seconds. Repeated IP blocks escalate: 30 minutes, then
    2 hours, then 24 hours. Upstream provider throttling surfaces as 503.
- name: Check domain availability
  scope: endpoint
  operationId: checkDomainsAvailability
  metric: requests_per_minute
  limit: 30
  timeFrame: minute
  secondary_limit:
    metric: requests_per_hour
    limit: 900
  description: 30 per minute or 900 per hour.
recoveryStrategies:
- name: Honor Retry-After
  description: On the OAuth init endpoints, wait the number of seconds in Retry-After before retrying.
- name: Exponential backoff
  description: >-
    Everywhere else there is no reset signal, so back off exponentially. The first-party SDKs
    auto-retry 429 and 5xx with linear backoff, maxRetries default 1.
- name: Batch and pace
  description: >-
    Instantly's guidance: batch bulk work (their example is 100 leads per batch with a 2 second
    pause between batches) and spread scheduled automations across 2-4 runs a day instead of one.
notes:
- Retries are not safe by default — Instantly ships no idempotency-key mechanism. See conventions/instantly-ai-conventions.yml.