LinkedIn Keyword Post Monitor

Find newly observed LinkedIn posts about your selected topics, keep the original source text and links, and avoid processing the same qualifying posts on every scan. Useful for GTM research, niche content discovery and finding questions worth answering.

Live collection uses the separately billed HarvestAPI LinkedIn Post Search. This Actor adds filtering, evidence and persistent monitoring; it is not an independent LinkedIn extraction engine. Existing-dataset mode processes posts you have already collected without starting another source run. No LinkedIn password, session cookies or developer-owned API key are included or requested.

Live mode requires an Apify plan that permits running public Store Actors. Some Creator plans restrict this; those accounts can use existing-dataset mode. The Actor does not change your subscription or silently use another account.

Quick start

{
  "sourceMode": "search",
  "queries": ["\"CRM migration\""],
  "postedLimit": "week",
  "sortBy": "date",
  "mode": "new",
  "monitorName": "crm-questions",
  "requestsOnly": true,
  "maxScanPosts": 25,
  "maxResults": 25,
  "maxSourceChargeUsd": 0.1,
  "sourceTimeoutSecs": 180
}

Run again with the same monitor name and search/filter settings to receive newly observed qualifying posts. Add an Apify schedule yourself if you want recurring scans; this Actor does not silently create schedules or publish comments.

requestsOnly uses simple English phrases and question punctuation. It can match rhetorical or promotional questions. Read the actual evidence before responding. Neither a matching term nor a question proves someone wants to buy your product.

For existing data, select Existing Apify dataset and enter a dataset ID in sourceDatasetId. The supported row shape includes id, linkedinUrl, content, optional author, postedAt and engagement. Source errors, non-post records and missing identities/text are not treated as valid posts. Dataset mode reads the first maxScanPosts rows; increase that cap to examine more rows. Search time-window options are not reapplied to existing datasets.

What you receive

• Original available post text, source URL and identity.
• Source publication date and author fields when available; unknown values stay null.
• Available engagement counts, without adding overlapping likes and reaction totals.
• Exact matched phrase text and character offsets, plus explicit request-language rule labels.
• A stable eventId, source-run IDs and field provenance.
• A separate SUMMARY containing coverage, source work, costs when available, pending output and errors.

Output is JSON and can be exported through Apify as CSV or Excel. Large nested fields remain available in JSON.

Prices and spending controls

Our monitoring fee is $1.98 per 1,000 delivered post records, plus a $0.00005 startup event per allocated GB of memory (minimum one event). Apify compute and storage usage are separate.

Live searches also pay HarvestAPI on the running customer's Apify account. On 1 October 2026 its base Free/Starter price was $2 per 1,000 collected posts, with tier discounts; an empty query can also incur a source event. Check the current source pricing. Profile enrichment, comments and reactions collection are disabled in our source input.

For example, collecting 100 posts and delivering 20 qualifying posts costs approximately $0.20 source fees + $0.0396 monitoring fees, plus applicable startup/platform costs. This is an illustration, not an all-in quote. Filtered-out posts and previously seen posts can still incur upstream collection charges. A run producing zero new posts can still cost money.

• maxScanPosts caps requested source posts across all queries; each query receives floor(maxScanPosts / queryCount).
• maxResults caps our delivered output; it does not cap the cost of collecting candidates.
• maxSourceChargeUsd is the separate child-run PPE ceiling. It is not the parent Actor or total account budget.
• Apify's maximum Actor charge controls our event fees, not every source/platform charge.
• sourceTimeoutSecs bounds collection and is reduced when the parent has less time remaining.

Monitoring behavior

The first new-post scan emits matching observations by default. Set emitInitial: false to seed a qualifying baseline without output fees. Later observations are compared with that saved baseline. Changes to reaction counts do not generate new-post events.

Queries, source selection, time window, ordering and filters are part of the history scope. Changing them creates a separate baseline. Changing maxResults does not reset history. Keep monitorName stable; change it deliberately to start fresh. History belongs to the running user and Actor.

Qualified overflow is saved before delivery. The next run drains that pending output before collecting new data. A monitor lock prevents simultaneous writers. If a crash occurs after output but before its checkpoint is saved, an event can be delivered and charged again with the same eventId; consumers should deduplicate this ID. Delivery is at least once, not exactly once.

History does not silently expire. It stops at the configured item limit or 16 MB size limit rather than forgetting identities and unexpectedly repeating old posts.

Coverage and troubleshooting

This is a bounded observation workflow. Search results can be ranked, incomplete, delayed or restricted by the source. “New” means newly observed in this monitor, not proof that a post was just published. Missing posts never generate deletion events. This Actor cannot promise all LinkedIn posts, global virality or private analytics.

Failed or partial source runs do not establish an empty baseline. Inspect SUMMARY and ERROR. SOURCE_RUN records source launch information; an uncertain launch is not automatically retried, preventing an accidental second charge inside the same parent run. Existing-dataset mode can be used to process a completed compatible source dataset.

Source access, platform permissions and pricing can change. The Actor fails before collection if the configured source no longer supports the expected limited-permission, pay-per-event contract.