How writes work
Queued actions, sender selection, dry runs and idempotency.
Writes are queued
LinkedIn work is slow and must be paced, so no write blocks. A write tool checks your request, saves it as an action and returns at once:
{
"action_id": "...",
"status": "queued",
"dry_run": false,
"sender_id": "...",
"run_after": "2026-12-01T09:14:00Z",
"deferred_reason": null
}Each sender has its own queue and runs one action at a time, with a random pause between actions.
Action statuses
| Status | Meaning |
|---|---|
scheduled | Waiting for its time or its condition (follow-ups). |
queued | Due, waiting for the sender's turn. |
dispatched / running | Being executed. |
succeeded | Done. |
failed | LinkedIn refused it, or it failed after retries. error says why. |
cancelled | You cancelled it with cancel_action. |
skipped | Its condition no longer held when it was due. |
Picking a sender
Write tools take an optional sender_id. Leave it out and SendCrew picks the healthiest warm sender that still has quota today. Two rules always win over your choice:
- A reply goes from the sender that owns the thread.
- A follow-up to a contact goes from the sender that first contacted them.
Over the cap is not an error
If a sender has used today's quota, the write is still accepted. It queues with a deferred_reason and a later run_after. Only a write that must never run (for example, a person another sender already contacted) is refused. See errors.
Dry runs
Every write accepts dry_run: true. The response has status would_queue and says which sender would send it and when, and saves nothing.
Retries and campaigns
idempotency_key: send the same key again and you get the first action back instead of a duplicate. Use it whenever your agent may retry.campaign: a free-text label. Actions with the same label are grouped in Campaigns and inget_stats.