Scripts and a prompt for analyzing competition in an Upwork niche. From my video on finding a niche as a freelancer.
The idea: instead of eyeballing freelancer profiles by hand, you extract their profile data, clean it up, and let an LLM break down the whole cohort for you — every contract, realized rates, who's a real specialist and who just slapped the keyword on their title.
| File | What it does |
|---|---|
grab.js |
Paste into the browser console on a freelancer's Upwork profile page. Copies the work history currently on screen to your clipboard. No scraping, no credentials — it reads what the page already loaded. |
slim-profile.mjs |
Cleans the copied dump into compact JSON in results/<name>.json. Raw dumps are 100KB+ of descriptions and junk — this keeps each profile small so 10-15 of them fit in the LLM's context and it stays in the smart zone. Re-running merges new contracts into the same file, so you collect work history one page at a time. |
prompt.md |
The prompt you send to the AI (I use Claude Code) once your results/ folder is full. Paste your niche into the placeholders and it just works. |
- Open a freelancer's profile on Upwork
- Paste
grab.jsinto the browser console - Run
node slim-profile.mjs— it reads your clipboard and saves the profile - (optional) For full better results and fuller picture, click to the next work history page, repeat steps 2-3 until the script prints
COMPLETE - Repeat for 10-15 freelancers in your niche
- Open Claude Code in this folder, paste
prompt.mdwith your niche and baseline rate filled in
I recommend doing the collection part by hand — it only takes a few minutes and you won't give off bot vibes.
Keep everything — scripts, results, generated reports — in one dedicated folder, and always start your AI sessions from that folder. You're not dependent on chat context (clear anytime, continue where you left off), you can go deeper on a niche you like later, and with Claude Code the project memory is tied to the folder, so it accumulates.