Skip to main content
Glama
google-jobs-scraper

google-jobs-scraper

Google Jobs Scraper (Python client + MCP server)

Get Google Jobs listings (the job box in Google Search) as JSON or CSV from Python, or as a tool for Claude, Cursor and other MCP clients: title, company, location, salary, posted date, the full job description and the real apply URLs.

python google_jobs.py "software engineer" --location "New York, NY" --date week

What this repository is, plainly: a small open-source client (about 150 lines, standard library only) and an MCP server. They do not scrape Google themselves. The scraping runs in a hosted Apify Actor, Google Jobs Scraper - Descriptions, Apply Links & Salary, which you call with your own Apify token and pay per unique job. The next two sections explain why there is no key-free script here.

Is there a Google Jobs API?

No. Google does not offer a public API for the job listings it shows in Search. To get them as data you either load the Google Jobs page yourself or use a service that does.

Related MCP server: mcp-jobs

Why a plain HTTP request does not work

We tested it. On 19 November 2026, from a home connection, one request with a normal Chrome user agent:

curl -s -L -A "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/130.0.0.0 Safari/537.36" \
  "https://www.google.com/search?q=software+engineer+New+York&udm=8&hl=en&gl=us" -o page.html

Result: a 94 KB page with 0 job listings. It is not a CAPTCHA and not a block of the IP address; the page only contains a redirect to /httpservice/retry/enablejs, Google's "turn on JavaScript" page. requests, urllib and curl all get that page, so a script of the "download the HTML and parse it" kind returns nothing.

The hosted Actor gets the full Jobs page through a search-results proxy and then handles the rest: Google gives about 10 jobs per request and no page 2, so it repeats each search with Google's own filters and merges the results, removes duplicates, and asks again when Google answers with a short page. On 19 November 2026 the default search ("software engineer", New York) returned 57 unique jobs from 9 requests in 84 seconds.

Quick start

Requires Python 3.8+ and an Apify account (a free plan works; you need its API token). Nothing to install for the script.

git clone https://github.com/datalumehq/google-jobs-scraper.git
cd google-jobs-scraper

# Windows: set APIFY_TOKEN=apify_api_...
export APIFY_TOKEN=apify_api_...
python google_jobs.py "registered nurse" --location "Houston, TX"

Get a token at console.apify.com/settings/integrations. Every run has a spending cap (--max-charge, default $0.25): the Actor stops when the cap is reached.

Options

Option

What it does

Example

query

Job title, skill or company (required). Several allowed

"data analyst" "data engineer"

--location

City, region or country. Repeatable; every query is searched in every location

--location "London"

--country

2-letter country Google searches from (default us)

--country gb

--language

Google language (default en). Use the local one outside English-speaking countries

--language de

--date

any, today, 3days, week or month

--date today

--type

fulltime, parttime, contract or internship (English searches)

--type contract

--remote

Remote jobs only (English searches)

--remote

--coverage

Requests per search: fast (3), standard (9), max (13). More requests = more unique jobs

--coverage max

--max-jobs

Stop the run after this many jobs (0 = no limit)

--max-jobs 50

--max-charge

Spending cap for the run in USD (default 0.25)

--max-charge 1

--csv FILE

Write CSV instead of printing JSON

--csv jobs.csv

Output

One object per unique job. A real row from 19 November 2026 (sample_output.json), with the long fields shortened (the description, highlights and URLs are cut with "…"; this job had 8 apply links, 3 shown):

{
  "title": "Sr Software Engineer - AI and Observability",
  "company": "Disney Entertainment & Sports LLC",
  "location": "New York, NY",
  "via": "Indeed",
  "postedAgo": "2 days ago",
  "schedule": "Full-time",
  "salary": "149K-199K a year",
  "salaryText": "149K-199K a year",
  "applyUrl": "https://www.linkedin.com/jobs/view/sr-software-engineer-ai-and-observability-at-the-walt-disney-company-4491705793?utm_c…",
  "applyLinks": [
    {
      "site": "Indeed",
      "url": "https://www.indeed.com/viewjob?jk=a1a02d95f912eaa9&utm_campaign=google_jobs_apply&utm_sour…",
      "direct": false,
      "titleMatch": null
    },
    {
      "site": "LinkedIn",
      "url": "https://www.linkedin.com/jobs/view/sr-software-engineer-ai-and-observability-at-the-walt-d…",
      "direct": false,
      "titleMatch": true
    },
    {
      "site": "ShowbizJobs",
      "url": "https://www.showbizjobs.com/jobs/disney-sr-software-engineer-ai-and-observability-in-new-y…",
      "direct": true,
      "titleMatch": true
    }
  ],
  "highlights": {
    "Qualifications": [
      "5+ years of applicable experience in backend development, with a proven track record of building and scaling AI-powered or data-driven applications",
      "…"
    ],
    "Benefits": [
      "Job Posting Statement The hiring range for this position in New York City, NY is $148,700 - $199,400 per year, and in Glendale, CA is $141,900 - $190,300 per ye …",
      "…"
    ],
    "Responsibilities": [
      "Innovation: We develop and implement groundbreaking products and techniques that shape industry norms and solve complex and distinctive technical problems",
      "…"
    ]
  },
  "description": "Job Requirements\n\nJob Description\n\nDisney Entertainment and ESPN Product & Technology\n\nTechnology is at the heart of Disney’s past, present, and future. Disney Entertainment and ESPN Product & Technology is a global organization of engineers, product developers, designers, technologists, data scientists, and more – all working to build and advance the technological backbone for Disney’s media busi …",
  "tags": [
    "2 days ago",
    "149K-199K a year",
    "Full-time",
    "Health insurance"
  ],
  "jobId": "pjSsvYmU8sFPWpW-AAAAAA==",
  "googleJobsUrl": "https://www.google.com/search?ibp=htl;jobs&q=software+engineer+New+York,+NY+full+time&htidocid=pjSsvYmU8sFPWpW-AAAAAA%3D…",
  "searchQuery": "software engineer",
  "searchLocation": "New York, NY",
  "country": "us",
  "language": "en",
  "scrapedAt": "2026-11-19T12:16:44+00:00"
}

--csv writes the flat columns: title, company, location, via, postedAgo, schedule, salary, applyUrl, jobId, searchQuery, searchLocation. Use JSON when you need the description, highlights or all apply links.

What it costs

The client and the MCP server are free and MIT licensed. The Actor is paid per unique job saved, billed by Apify to your own account; the current price is on the Actor page. Searches with no results, pages Google blocked and duplicate jobs are not charged.

Use it as an MCP server

mcp_server.py exposes three tools to any MCP client (Claude Desktop, Claude Code, Cursor, VS Code). Requires Python 3.10+.

Tool

Cost

What it returns

google_jobs_link

Free, no API key, no network request

The Google Jobs page URL for a search, for you to open in a browser. A link only, no job data

start_job_search

Paid per unique job through your Apify account (price on the Actor page)

Starts a search and returns a run_id at once. Filters: date posted, employment type, remote. Each search has a spending cap (max_total_charge_usd, default $0.25) and a job limit (max_jobs, default 20)

get_job_results

Free (reads the results of a search you started)

The listings as data once the search has finished: title, company, location, salary, posted date, full description, apply URLs. While the search is still running it says so and the client calls it again

Why two steps: a search takes about a minute (59 seconds in our 19 November 2026 test with fast coverage), and some MCP clients stop waiting for a single tool call before that. Starting the search and fetching the results separately keeps every call short.

Install:

git clone https://github.com/datalumehq/google-jobs-scraper.git
cd google-jobs-scraper
pip install -r requirements.txt

Claude Desktop (claude_desktop_config.json), Cursor (~/.cursor/mcp.json) and most other clients use the same block. Use the full path to mcp_server.py:

{
  "mcpServers": {
    "google-jobs": {
      "command": "python",
      "args": ["/path/to/google-jobs-scraper/mcp_server.py"],
      "env": { "APIFY_TOKEN": "<your Apify token, needed for start_job_search and get_job_results>" }
    }
  }
}

Claude Code:

claude mcp add google-jobs -e APIFY_TOKEN=<your token> -- python /path/to/google-jobs-scraper/mcp_server.py

Then ask in plain language, for example:

  • "Find remote software engineer jobs posted today in the US and list the salaries"

  • "Which companies are hiring registered nurses in Houston this week? Give me the apply links"

Without APIFY_TOKEN the link tool works and the other two return a message explaining how to add a token.

Prefer not to run anything locally? Apify's hosted MCP endpoint serves the same Actor: https://mcp.apify.com?tools=datalume/google-jobs-scraper (sign in with Apify on first use).

Limits

  • No Apify token, no job data: this repository cannot read Google Jobs on its own (see above).

  • Google Jobs is not served in every country; searches there return nothing and cost nothing.

  • Salary is present only when Google shows pay for the job.

  • Employment type and remote filters use Google's English filter phrases. In other languages use --date and put the job type in the query.

  • Google decides what it shows. Results can differ from what you see in your own browser.

FAQ

Is scraping Google Jobs allowed? The listings are public pages that Google shows to anyone without logging in. The Actor saves only the public job post, does not log in anywhere and collects no personal data. Check the terms and laws that apply to your use.

Can I get more than 10 jobs per search? Yes. One request returns about 10 jobs; --coverage standard makes 9 requests with different Google filters and merges them (57 unique jobs in the 19 November 2026 test above). Niche searches return fewer.

Can I run it daily for new jobs only? Yes: run it with --date today and the same queries each day. jobId lets you skip jobs you already have.

Can I use requests/BeautifulSoup or Selenium instead? Plain HTTP libraries get the JavaScript page described above. A real browser can load Google Jobs, but that is a different kind of project (browser automation, proxies, blocks) and is not what this repository does.

Something is missing or wrong? Please open an issue here.

More

License

MIT. Built with AI assistance; numbers come from real runs on the dates shown.

Not affiliated with, endorsed by or sponsored by Google. This project only reads publicly available job listings shown by Google.

Available Tools

3 tools
get_job_resultsA

Get the jobs found by a search started with start_job_search. Step 2 of 2. Reading results is free; it never starts a new search.

Waits up to wait_secs for the search to finish. Returns {"finished", "status", "jobs",
"message"}. If "finished" is false the search is still running and "jobs" is empty:
call get_job_results again with the same run_id (do not start a new search).

Args:
    run_id: the run_id returned by start_job_search.
    wait_secs: how long to wait for the search to finish, 0-45 seconds.

Inputs: run_id (required), wait_secs.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4.6/5.0

    Scored across 3 tools

    Disambiguation5/5

    Each tool maps to a distinct step with no overlap: building a free browser link, starting a paid async search, and polling for results. Descriptions explicitly disambiguate the risky pair (google_jobs_link vs start_job_search) by telling the agent which to use when listings are needed as data.

    Naming Consistency4/5

    start_job_search and get_job_results follow a clean verb_noun pattern, and the two are explicitly labeled step 1/step 2. google_jobs_link deviates with a noun-style name rather than a verb, but it is still readable and semantically clear.

    Tool Count5/5

    Three tools is exactly the minimum surface for an async start/poll scraper plus a free link alternative. Each tool has a distinct role and none feels redundant or missing for this scope.

    Completeness4/5

    The start → poll lifecycle is fully covered, including wait_secs and cost caps, so an agent cannot hit a dead end while retrieving results. Minor gaps exist: no way to cancel an in-flight paid run or list prior runs, so stale run_ids must be tracked externally.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      Not graded
      quality
      B
      maintenance
      Enables AI assistants to search Google Jobs and retrieve structured job listings, including titles, companies, locations, descriptions, and direct apply links, through natural language queries. Supports location targeting, country/language filters, and pagination.
      -
    • A
      license
      Not graded
      quality
      C
      maintenance
      Enables natural-language job search and aggregation from multiple recruitment websites with zero configuration, providing filtered results and standardized output for AI assistants.
      1
      ISC
    • F
      license
      Not graded
      quality
      A
      maintenance
      Enables AI agents to search, filter, and apply across nine job boards in eight countries — including ~4,700 salary-transparent tech jobs plus Swiss and Singapore listings — with results returned as JSON data and applied jobs tracked so duplicates never resurface.
      -
    • F
      license
      Not graded
      quality
      A
      maintenance
      Enables AI agents to search, filter, and apply to salary-transparent tech jobs across nine international job boards, including Switzerland's jobs.ch and jobup.ch and Singapore's MyCareersFuture. Results are returned as JSON with salary, remote, tech, and posting-date filters, and previously applied postings are remembered so they drop out of subsequent listings.
      -