LinkedIn Jobs Scraper — no login, strict matching, limits that hold

Search LinkedIn job postings and get clean, stable rows: title, company, location, posted date, salary when shown, and the full description on request. No LinkedIn account, no cookies, no session to get banned — this Actor reads the same public listings LinkedIn serves to any logged-out visitor.

Why this one

Job scrapers are the most-complained-about category on the store, and the complaints are always the same three:

1. "One week it matches strictly, another week 5k+ irrelevant results." Here, titleMustInclude / titleMustExclude are enforced on every row, in one place in the code, every run. A job whose title does not match is never delivered and never charged for.
2. "I set a limit and still paid for thousands." maxJobs and maxCostUsd are hard ceilings checked before each fetch. The run stops collecting at the ceiling — it does not keep spending and trim the output afterwards.
3. "The output format changed and broke my pipeline." Every row carries schemaVersion. Fields get added, never renamed or removed, without that number changing.

What you get

One row per job:

Field	Example
title	Data Engineer
company / companyUrl	Acme Corp / linkedin.com/company/acme
location	New York, NY
postedAt	2026-08-20
salary	$140,000 – $180,000 (when LinkedIn shows it)
jobUrl	linkedin.com/jobs/view/4012345678
schemaVersion	1

With Fetch full descriptions on, each row also carries the complete description (text and HTML), seniority level, employment type, job function, industries, and the applicant count when shown.

Every run writes a RUN_SUMMARY saying exactly what was delivered, what your filters rejected and why, what failed, and what it cost against your cap.

Filters that actually filter

• Posted since — checked against each posting's own date, per row. Jobs outside the range are not delivered and not charged.
• Job type, experience level, workplace (on-site / remote / hybrid) — LinkedIn's own search filters, passed through.
• Title must include / exclude — plain, case-insensitive, predictable substring matching applied after the search returns. This is the tool for pipelines that need "data engineer" to mean data engineer.

Honest limits

LinkedIn's public guest search serves at most 1,000 results per query. That is LinkedIn's ceiling, not ours, and the Actor says so in the log when you hit it instead of pretending the search was exhausted. To reach deeper, narrow the query: split by location, date range, or job type.

Typical uses

• Feed an ATS or CRM with fresh postings for target roles, on a schedule
• Track which companies are hiring for a role, week over week
• Salary and demand research by title and region
• Lead generation: companies hiring for X are buying tools for X

Pricing

Pay per result: one price per job row, one small extra price per full description you ask for. No subscription, no minimum. Set Max spend (USD) and the run cannot cost more than that number.

FAQ

Do I need a LinkedIn account? Can my account get banned? No account is used at all. The Actor reads the public guest listings LinkedIn serves to logged-out visitors, so there is nothing to ban.

How many jobs can one search return? LinkedIn serves at most 1,000 guest results per query — that is LinkedIn's ceiling, and the run log says so when you reach it. To go deeper, split the search by location, date range or job type and combine the runs.

Does it return salaries? When LinkedIn shows one on the listing, yes, in the salary field. Most postings do not carry one; the field is null rather than a guess.

Do repeated jobs get charged twice? Not within a run — promoted postings repeat across pages and are deduplicated by job id before anything is charged.

Can I export to CSV or feed my ATS? Yes — every run downloads as CSV, Excel, JSON or JSONL, or pull the dataset via the API. schemaVersion on each row means the columns will not move under your pipeline.

Can I run it daily? Yes. Save your search as a Task, schedule it, and pair it with postedSince so each run returns only what is new since the last one.