MCP Job Search

# LinkedIn Job Search Programmatic Scraping Guide ## 1. Quick Reference of URL Parameters | Parameter | Purpose | Example | Notes | |-----------|---------|---------|-------| | `keywords` | Search string (supports Boolean) | `keywords=python%20developer` | See Section&nbsp;4 for Boolean syntax | | `location` | Plain‑text location (city/state/country) | `location=Sydney%2C%20New%20South%20Wales%2C%20Australia` | Optional if `geoId` present | | `geoId` | LinkedIn geographic ID | `geoId=102026833` | Disambiguates location text | | `f_PP` | Primary place filter (city/region IDs) | `f_PP=102026833%2C103644278` | Comma‑separated list of IDs (OR logic) | | `distance` | Radius in miles (for city searches) | `distance=25` | Ignored for country‑level filters | | `f_TPR` | Time posted (seconds) | `f_TPR=r86400` | `r86400` = 24 h, `r3600` = 1 h | | `f_E` | Experience level | `f_E=2` | 1 Intern, 2 Entry, 3 Assoc, 4 Sr, 5 Dir, 6 Exec | | `f_JT` | Job type | `f_JT=F,P` | F Full‑time, P Part‑time, C Contract, T Temp, V Vol, I Intern, O Other | | `f_WT` | Workplace (on‑site/remote/hybrid) | `f_WT=1,3` | 1 On‑site, 2 Remote, 3 Hybrid | | `f_AL` | Easy Apply only | `f_AL=true` | | | `f_JIYN` | Fewer than 10 applicants | `f_JIYN=true` | | | `sortBy` | Sort order | `sortBy=DD` | DD = newest, R = relevance | | `start` | Pagination offset | `start=25` | 25 results per page | --- ## 2. Location Filtering Deep Dive ### 2.1 How job posters define location * **On‑site & Hybrid:** must pick a **city/metro**. * **Remote:** may pick city, state, or country. * Location cannot be edited once posted. ### 2.2 How searches interpret location * **City‑level filter** → returns only jobs tagged with *that* city. * **Country‑level filter** → returns jobs tagged with any city in that country **plus** posts tagged only with the country itself. * A job tagged “Australia” will **not** appear in a city‑only search for “Sydney”. ### 2.3 Recommended parameters | Scenario | Recommended Params | |----------|--------------------| | Specific city | `location=<city>&geoId=<id>` **or** `f_PP=<id>` | | Multiple cities | `f_PP=<id1>,<id2>,<id3>` | | Country | `location=<country>&geoId=<id>` | | Remote only (anywhere) | `location=Remote` **or** `f_WT=2` | | Remote within region | `location=<country>&geoId=<id>&f_WT=2` | | Hybrid only | `f_WT=3` | ### 2.4 Getting geo IDs programmatically ``` GET https://www.linkedin.com/jobs-guest/api/typeaheadHits ?origin=jserp&typeaheadType=GEO&geoTypes=POPULATED_PLACE &query=Sydney ``` Parse the `urn:li:fs_geo:<id>` values and store for use in `geoId` or `f_PP`. --- ## 3. Example URLs ```text # City with 25 mi radius https://www.linkedin.com/jobs/search/?keywords=software%20engineer&location=Sydney,%20New%20South%20Wales,%20Australia&geoId=102026833&distance=25&sortBy=DD # Sydney OR Melbourne https://www.linkedin.com/jobs/search/?keywords=frontend%20developer&f_PP=102026833%2C103884924&sortBy=DD # Remote jobs in Australia https://www.linkedin.com/jobs/search/?keywords=data%20scientist&location=Australia&geoId=101452733&f_WT=2 # Last‑hour postings in US https://www.linkedin.com/jobs/search/?keywords=golang&location=United%20States&geoId=103644278&f_TPR=r3600 ``` --- ## 4. Keyword Search Behaviour LinkedIn supports Boolean logic inside the `keywords` parameter. | Operator (caps) | Meaning | Example | URL‑encoded | |-----------------|---------|---------|-------------| | _(space)_ | **OR** (default) | `marketing manager` | `marketing%20manager` | | `AND` | All terms must appear | `python AND rust` | `python%20AND%20rust` | | `OR` | Any term may appear | `developer OR engineer` | `developer%20OR%20engineer` | | `NOT` | Exclude term | `javascript NOT react` | `javascript%20NOT%20react` | | `" "` | Exact phrase | `"machine learning"` | `%22machine%20learning%22` | | `( )` | Grouping | `("machine learning" OR AI) AND python` | `%28%22machine%20learning%22%20OR%20AI%29%20AND%20python` | **Tips** * Default space acts as **OR** — use `AND` to force both. * Quote multi‑word titles: `"site reliability engineer"`. * Combine stacks: `backend AND ("node.js" OR node)`. * Use `NOT` to weed out noise: `analyst NOT "business analyst"`. * Always URL‑encode spaces (`%20`) and symbols. --- ## 5. Scraper Implementation Advice * **Prefer direct URL construction** once you know `geoId`/`f_PP` – faster than UI clicks. * **UI automation fallback:** if LinkedIn changes param keys, click filters in Playwright, watch URL diff, copy new param. * **Pagination:** increment `start=0,25,50…` until no results or limit reached. * **Login & anti‑bot:** keep real session cookies, randomize waits, rotate user‑agents. * **Edge cases:** * Run a country‑level remote query to capture remote‑only jobs missed by city filters. * Hybrid jobs match their specified city. * Multi‑location postings are rare; covering each desired region individually is safest. --- ## 6. Quick Cheatsheet ```text # Sydney, last 24 h, <10 applicants, entry‑or‑assoc https://www.linkedin.com/jobs/search/?keywords=python%20developer&location=Sydney,%20NSW,%20Australia&geoId=102026833&f_TPR=r86400&f_E=2%2C3&f_JIYN=true&sortBy=DD # Remote US, data + ML, full‑time https://www.linkedin.com/jobs/search/?keywords=(data%20scientist%20OR%20"machine%20learning")%20AND%20python&location=United%20States&geoId=103644278&f_WT=2&f_JT=F ```