Rate limits
How many requests per minute and per second, what each operation costs, and how to react to a 429.
Limits protect the platform. They are counted per account (per agency for an agency connection), all the account's keys and connections together, over two windows: the minute, and the second (the burst). Test mode has its own counters: your trials take nothing from the live budget. The AI connector has its own too.
| Plan | Per minute | Per second |
|---|---|---|
| Essential | 120 | 10 |
| Pro | 300 | 20 |
| Team (per seat) | 300 | 20 |
| Agency connection (all its accounts) | 1000 | 40 |
What an operation costs
A request costs 1, except those that make the platform work harder:
| Operation | Cost |
|---|---|
POST /v1/imports | 10 |
POST /v1/closer/agent/test | 5 |
POST /v1/closer/knowledge | 5 |
POST /v1/webhook_endpoints/{id}/deliveries/{delivery_id}/retry | 5 |
POST /v1/webhook_endpoints/{id}/test | 5 |
PUT /v1/prospects | 3 |
POST /v1/campaigns/{id}/preview_message | 2 |
POST /v1/campaigns/{id}/reanalyze_website | 2 |
POST /v1/campaigns/preview_target | 2 |
GET /v1/conversations | 2 |
GET /v1/conversations/search | 2 |
POST /v1/exports | 2 |
POST /v1/onboarding/preview_prospects | 2 |
POST /v1/people/search | 2 |
POST /v1/prospects/search | 2 |
The headers
Every response tells you where you stand:
RateLimit-Policy: the two windows, for example"minute";q=300;w=60, "burst";q=20;w=1.RateLimit: what is left in the minute and the seconds before it resets, for example"minute";r=287;t=42.X-RateLimit-Limit,X-RateLimit-Remaining,X-RateLimit-Reset: the same values, for tools that read these names (X-RateLimit-Resetcounts seconds, not a date).Retry-After, on a refusal: the seconds to wait.- On a refusal by the per-second limit, these headers describe the one-second window:
RateLimit: "burst";r=0;t=1,X-RateLimit-Limitis the per-second limit,Retry-After: 1.
Beyond that: 429
Beyond that, the request is refused with 429 rate_limited and Retry-After, and nothing is done. Wait for the delay, then retry: the SDKs do it for you. Read in pages of 100 and follow changes with updated_since or webhooks rather than reading everything in a loop.
Sending limits are something else
The number of invitations and messages per day, the sending hours and the pauses that protect the LinkedIn account never go through a 429: a send beyond them is scheduled (202, status: "scheduled", scheduled_at) or refused with a code that says so (daily_cap_reached, sends_paused). See how Lidmeo works.