generated: '2026-07-20'
method: searched
source: https://docs.millimetric.ai/reference/rate-limits.md
description: >-
  Rate limits are per project, per route, enforced by an in-memory token bucket on the
  Cloudflare Worker. There is no per-key or per-IP limit on top. Read endpoints are
  unlimited. On exhaustion the Worker returns 429 rate_limited with a Retry-After header.
scope: per-project, per-route
enforcement: in-memory token bucket (Cloudflare Worker)
signal:
  status: 429
  code: rate_limited
  header: 'Retry-After: <seconds>'
  body_field: retry_after_s
rate_limits:
  - endpoint: POST /v1/track
    refill_rate: 50/sec
    burst_capacity: 200
    limit_count: 50
    window: 1s
  - endpoint: POST /v1/batch
    refill_rate: 5/sec
    burst_capacity: 20
    limit_count: 5
    window: 1s
    note: Each call delivers up to 1000 events (5/sec x 1000 = 5,000 events/sec sustained).
  - endpoint: GET /v1/query
    refill_rate: unlimited
  - endpoint: GET /v1/stats
    refill_rate: unlimited
  - endpoint: GET /v1/sources
    refill_rate: unlimited
  - endpoint: POST /v1/identify
    refill_rate: unlimited
  - endpoint: POST /v1/forget
    refill_rate: unlimited
  - endpoint: POST /mcp
    refill_rate: inherits the underlying tool's limit
guidance:
  - Use the browser SDK (batches every 20 events or 2s) or Node SDK (flushAt) to stay under limits.
  - Backfill via /v1/batch in chunks of 1000 with ~200ms spacing (~5/sec).
  - Limits are per project; spread traffic across projects for more headroom.