Naukri Job Hunter

Find remote jobs and side gigs (part-time, freelance, contract) on naukri.com, score every job against your resume, get a ranked HTML report, and optionally auto-apply. Works as a one-command script or as an AI agent (MCP server for Cursor, Claude Desktop and other MCP clients).

• One command: ./run.sh; a first-run wizard sets everything up

• Resume-based match score (0-100) with matched skills per job

• Two hunts: regular remote jobs, and side gigs you can do alongside a job

• Only shows new jobs each run; HTML + CSV reports

• Choose how to apply: auto-apply, pick jobs from a numbered list, or get a contact list (recruiter emails, phones, company links, prefilled email drafts) to send your resume yourself

• Auto-apply has a dry run, confirmation and daily caps

• Your resume, login and results stay on your machine

Quick start

git clone https://github.com/psflwork/naukri-job-hunter.git
cd naukri-job-hunter
./run.sh

The first run:

1. creates a Python environment and installs dependencies,

2. asks for your resume PDF (drag the file into the terminal), years of experience and the roles you want,

3. asks for your Naukri email/password (optional; saved only locally in .env), then

4. opens Chrome, searches Naukri, scores the jobs and opens the report, then shows a menu:

What next?
  [a] Auto-apply to 4 eligible jobs (of top 10)
  [p] Pick jobs to apply to
  [c] Contact list: emails / phones / links to send your resume manually
  [o] Open top 10 jobs in your browser
  [q] Quit

After that, just run ./run.sh (regular jobs) or ./run.sh side (side gigs) whenever you want.

Requirements: Python 3.10+ and Google Chrome. macOS and Linux work out of the box; on Windows use WSL or Git Bash.

Run it as ./run.sh from the project folder (with the ./). It uses its own Python in .venv, so you never need to activate anything.

Related MCP server: remoote-mcp

Usage

./run.sh                 # regular hunt: search, report, then the menu above
./run.sh side            # side-gig hunt (part-time / freelance / contract)
./run.sh --no-apply      # search + report only
./run.sh --apply         # auto-apply without the menu (for scheduled runs)
./run.sh --all           # include jobs already seen in earlier runs
./run.sh --limit 10      # only the 10 best matches this run (default: max_results in config)
./run.sh --top 20        # consider the top 20 matches in the menu
./run.sh setup           # re-run the setup wizard

Shortcuts

Work on the latest search results; put side first for the gig hunt (./run.sh side pick).

The contact list only shows emails/phones that recruiters actually published (Naukri hides recruiter details otherwise); it never guesses addresses. Edit the email template under outreach in your config.

Search every morning at 9:00 without applying (crontab -e):

0 9 * * * cd /path/to/naukri-job-hunter && ./run.sh --no-apply --no-open >> output/cron.log 2>&1

How it works

flowchart LR
    You([You]) -->|"chat: find remote jobs"| Agent[AI agent<br/>Cursor / Claude]
    You -->|./run.sh| CLI[run.sh → hunt.py]
    Agent -->|MCP tools| MCP[mcp_server.py]
    MCP --> Core
    CLI --> Core

    subgraph Core[Core]
        Search[run_search] --> Matcher[matcher.py<br/>score 0-100]
        Naukri[naukri.py<br/>Playwright + Chrome]
    end

    Search --> Naukri
    Naukri -->|visible Chrome window| Site[(naukri.com<br/>jobapi/v3/search<br/>jobapi/v4/job)]
    Resume[/your resume PDF/] --> Matcher
    Config[/config.yaml/] --> Search
    Env[/.env credentials/] --> Naukri

    Matcher --> Out[output/*.html + *.csv]
    Matcher --> State[(data/<br/>seen, applied,<br/>latest results,<br/>browser session)]

1. Search: for each query in the config, a real Chrome window opens Naukri's search page with the Remote, experience and posted-in-last-N-days filters. The script reads the JSON the page itself loads, so it doesn't depend on the page's HTML layout.

2. Filter: drops excluded titles/companies, jobs already applied to, and jobs seen in earlier runs.

3. Score: each job gets 0-100 against your resume:

   Part	Points	Based on
   Skills	50	skills from config found in the job, plus job tags found in your resume
   Title	25	strong_titles / target_titles in the job title
   Experience	15	your years inside the job's min-max range
   Remote	10	location says Remote

4. Report: jobs at or above min_score go to output/jobs_<timestamp>.html and .csv.

5. Apply (optional): clicks Naukri's own Apply button. Company-site jobs and jobs with recruiter questionnaires are left for you to apply manually.

Two hunts: regular jobs and side gigs

Each hunt keeps its own "already seen" history and report (output/side_jobs_*.html); applied jobs are shared, so you never apply twice. Add more hunts by creating config.<name>.yaml and running ./run.sh <name>.

Naukri has no part-time filter, so the side hunt works differently:

flowchart LR
    Q["Gig keywords<br/>freelance, part time,<br/>contract developer..."] --> S[Naukri search<br/>remote, last 30 days]
    S --> G{Gig signal?<br/>part-time, freelance,<br/>6 months contract,<br/>secondary income...}
    G -- no --> X[dropped]
    G -- yes --> F{Tech role and<br/>not junior?}
    F -- no --> X
    F -- yes --> Sc["Score<br/>skills 45, gig 20,<br/>title 15, exp 10, remote 10"]
    Sc --> R[side_jobs report<br/>with Gig signals column]

• Gig signal required: phrases like part-time, freelance, contractual, 6 months contract, secondary income, hourly in the job. A bare contract/weekend/consultant only counts in the title, and noise such as contract testing or smart contracts is ignored.

• Filters: drops non-tech titles (teacher, sales, accountant, ...), jobs with none of your skills, and junior gigs (experience range below 5 years).

AI agent (MCP)

mcp_server.py exposes the hunter as MCP tools, so an AI assistant can search, read job details, shortlist with reasons, write a tailored pitch, and apply only to the jobs you approve.

Cursor: the server and an agent rule (.cursor/rules/naukri-job-hunter.mdc) are already set up in this repo. Open the folder in Cursor, enable naukri-job-hunter under Settings → MCP, then ask e.g. "Find remote senior developer jobs from the last 3 days and shortlist the best 5" or "Find part-time or freelance gigs I can do alongside my job."

Claude Desktop / other MCP clients: run ./run.sh setup once, then add:

{
  "mcpServers": {
    "naukri-job-hunter": {
      "command": "/path/to/naukri-job-hunter/.venv/bin/python",
      "args": ["/path/to/naukri-job-hunter/mcp_server.py"]
    }
  }
}

sequenceDiagram
    actor You
    participant A as AI agent
    participant M as MCP server
    participant N as naukri.com

    You->>A: "Find remote engineering jobs"
    A->>M: get_candidate_profile
    M-->>A: resume text + preferences
    A->>M: search_jobs
    M->>N: search pages (logs in from .env if needed)
    N-->>M: job JSON
    M-->>A: ranked matches + report path
    loop top 5-8 jobs
        A->>M: get_job_details(job_id)
        M->>N: job page
        M-->>A: full JD, skills, applicants
    end
    A-->>You: shortlist with fit + tailored pitch
    You->>A: "Apply to #1 and #3"
    A->>M: apply_to_job(job_id, confirm=true)
    M->>N: click Apply
    M-->>A: applied / needs_manual
    A-->>You: results

Configuration

./run.sh setup creates config.yaml and config.side.yaml from examples/ and fills in your resume, experience and roles. Edit them any time to tune results:

Privacy

Everything runs locally. These files are git-ignored and never leave your machine: your resume (*.pdf), Naukri login (.env, readable only by you), your configs, the saved browser session and job history (data/), and reports (output/).

Troubleshooting

Project layout

Disclaimer

This is an unofficial personal-productivity tool, not affiliated with Naukri.com / Info Edge. Naukri's terms don't allow automated use; searching with polite delays is low risk, but heavy auto-applying can get an account flagged. Keep the caps low, review jobs before applying, and use it at your own risk.

License

MIT