Upwork Jobs Scraper

Yes — a brief posted in public is public data. Everything we collect is what any visitor can read on Upwork without signing in. More in our guide to the legality of web scraping.

No login required. No account to sync, no cookie to paste, no Chrome extension — and no Upwork account of yours to put at risk. Paste a search and go, which is rarer than it sounds on a marketplace this well guarded.

50 data points per job. Title, the full brief, when it went live, fixed budget or hourly range, duration, weekly workload, experience level, the complete skill list, how many freelancers are wanted — and the client behind it when Job Details is on.

About 800 jobs a minute. Switch on Job Details and it settles at roughly 10 a minute — that is the price of the client block. Keep it for shortlists, leave it off when you are sweeping a whole category.

Wherever you already work. Google Sheets, Make, n8n, Amazon S3, email, webhooks — plus our MCP server, so Claude or ChatGPT can launch a run and read the results itself. CSV, Excel and JSON too. See all integrations.

1 extra credit per job, and it roughly doubles the export. You get the client behind the brief — country, total spent, rating, reviews, hires, payment verification, member since — plus the exact proposal count and the applicant requirements. 31 columns become 64.

Yes — and the jobs you filter out cost you nothing. Category, hourly or fixed, budget range in dollars, experience level, the client's hiring history, the client's country and how recently it was posted. Whatever your URL already sets is left alone.

Yes, in fields you can actually sort on. Fixed-price jobs bring the budget amount, hourly jobs a minimum and a maximum rate, and every job carries the figure Upwork displays, the currency and whether it is hourly or fixed.

Yes — set a Newer Than threshold. Give it 24h, 7d, 2w or an exact date in your own timezone, and the search flips to newest-first and stops the moment it reaches older jobs. A daily re-run of the same search costs almost nothing.

All three work. An Upwork search URL with your filters, a plain keyword like python developer, or the link to one posting if you only want that job. Feed it a single job link and Job Details comes free.

URL, such as https://www.upwork.com/jobs/~022089069371186411351. Paste thousands at once or upload a CSV; duplicate inputs are skipped. 12 settings shape the run: see them in the playground.

50 data attributes per job, each with its description and an example: see the full list.

No. Duplicates are removed, and you only pay for unique results.

As fresh as your run: each run collects the data from the site at the moment it runs, nothing comes from a stored database.

CSV, Excel or JSON from the dashboard, through the API, or sent after every run to Google Sheets, Amazon S3, an SFTP server, a webhook or by email.

7 days on the free plan, 28 days on paid plans. Then they are deleted. Download or deliver what you want to keep.

Around 800 jobs a minute for one squid. Your plan's slots run several squids at the same time.

100.00% of its runs had no incident over the last 90 days (0 incidents, 100% fixed). Every day is on this page.

Our team fixes the scraper: every scraper on lobstr.io is built and maintained by us. How we measure.

Yes, on paid plans: every hour, every day, every weekday or on a custom schedule, in your time zone.

Yes: set the most results per task and per run in the settings. The run stops there and so does the bill.

1 credit per result. Options add to it: Job Details +1. The playground shows the price as you pick them.

Yes. The free plan gives 100 credits every month, 100 results with this scraper, with exports capped at 30 rows.

Every month a plan gives credits, which pay for results, and slots, which set how many squids run at the same time. Starter: $20 for 10,000 credits; Pro: $100 for 100,000 credits; Team: $500 for 1,000,000 credits. Upgrades start right away and are prorated; downgrades and cancellations take effect at the end of the cycle.

No. Your credits refresh every month with your plan.

You choose, per squid: pause and pick up where it stopped, or end the run. To keep going, upgrade or reset your billing cycle: your plan renews right away and your credits come back. Auto Billing Reset does it for you when your usage crosses a threshold you set.

Yes: paying yearly takes 20% off every plan.

A monthly plan: within 10 days of the charge, if no credit was used in that period. A yearly plan: within 30 days, in full if no credit was used, otherwise minus one month. Cancel from your billing page, then write to contact@lobstr.io. The details are in the terms of use.

Collecting publicly available data is generally allowed, but you choose what you collect and you are responsible for how you use it. Our terms of use (article 6) ask every user to respect the rights of the sites, not to extract substantial parts of a database, not to access confidential data and to follow personal data law such as the GDPR. This is not legal advice.

When results contain personal data, you are the data controller and lobstr.io, a French company, processes them on your behalf, under our personal data agreement. You need a lawful basis for what you collect, must inform the people concerned and honor their requests.

Results: 7 days on the free plan, 28 days on paid plans, then they are deleted; you can delete them sooner. Your account data: see the privacy notice.

Yes, through the REST API at https://api.lobstr.io/v1, with your key from the dashboard. The playground’s Code tab writes the calls for your settings, in cURL and Python. The documentation has the full reference.

Yes. Add our MCP server, https://mcp.lobstr.io/mcp, to Claude, ChatGPT, Cursor or any MCP client and sign in with lobstr.io. Your agent shows the cost and asks you before a run spends credits. The playground’s Agent tab gives you a prompt.

No. Credits pay for results, not for requests.

Yes. Point a webhook at your server and lobstr.io calls it when a run succeeds or fails.

Yes, per endpoint: 120 requests a minute on squids and runs, 90 on tasks, 2 a second on results. Above that the API answers 429 with a Retry-After header.