LinkedIn API v2 search pagination: bounding limit and cursor lifetime for a large dataset export

1. Maximum page size (limit)

The v2 search endpoints accept a limit parameter with a hard ceiling of 100 records per request. Sending a value greater than 100 returns a 400 Bad Request with error code INVALID_PARAMETER; the service does not silently truncate the response.

2. Cursor lifetime and checkpoint strategy

Pagination cursors (found in paging.next or paging.cursor) remain valid for roughly 5 minutes. After expiry the same cursor yields a 400 with code EXPIRED_CURSOR. Because the TTL is not covered by an SLA, treat it as a best‑effort window.

Recommended checkpoint pattern for a resumable export:

1. After each successful page fetch, persist the returned cursor value (e.g., write to a durable store).
2. When the job restarts, resume with the stored cursor if the elapsed time since the last fetch is < 5 minutes.
3. If the cursor has expired, re‑issue the original search request (without a cursor) and, if needed, apply a start offset that you have tracked separately to skip already‑processed records.

3. Offset‑based paging (start) reliability

The start parameter is not supported on v2 search endpoints. Supplying start is ignored and the response always returns the first page. Cursor‑based paging is the only supported mechanism for walking large result sets.

Verification steps you can run now

# 1. Confirm limit ceiling
curl -G "https://api.linkedin.com/v2/search" \
  -d q=people -d limit=101 \
  -H "Authorization: Bearer $TOKEN"
# Expect 400 INVALID_PARAMETER

# 2. Observe cursor expiry
curl -G "https://api.linkedin.com/v2/search" \
  -d q=people -d limit=10 \
  -H "Authorization: Bearer $TOKEN" > page1.json
CURSOR=$(jq -r '.paging.cursor' page1.json)
sleep 360   # 6 minutes
curl -G "https://api.linkedin.com/v2/search" \
  -d q=people -d limit=10 -d cursor="$CURSOR" \
  -H "Authorization: Bearer $TOKEN"
# Expect 400 EXPIRED_CURSOR

# 3. Test offset ignorance
curl -G "https://api.linkedin.com/v2/search" \
  -d q=people -d limit=10 -d start=50 \
  -H "Authorization: Bearer $TOKEN"
# Response should match first page (start ignored)

Assumptions & uncertainties

• Values reflect the public LinkedIn Marketing Developer Platform documentation as of 2026‑10; partner‑tier contracts may have different caps.
• Cursor TTL is approximate; network latency or index refreshes can invalidate cursors sooner.
• Result ordering is not guaranteed stable across pages; duplicates or gaps can appear if the index changes during the walk.
• The count field in the paging object is an estimate capped at 10 000 and should not drive progress reporting.

One diagnostic detail that would change the recommendation

If your application holds a partner‑level API contract that documents a higher limit ceiling or a longer cursor TTL, the above bounds would need to be adjusted. Please confirm the exact tier you are using.