jobhunt_mcp

A personal MCP server that turns Claude into your job-search operations manager:

Search live postings across LinkedIn, Indeed, ZipRecruiter, Glassdoor (JSearch API) plus Adzuna
Track every application in a Google Sheet (status, priority, notes)
Follow up on schedule — overdue nudges surface automatically
Network — keep contacts per company and pull them into prep
Tailor — one tool assembles the job description + your master resume + contacts so Claude can write a targeted resume and cover letter, saved straight to Google Drive

By design it never auto-submits applications. Auto-submission bots violate job-board terms of service and get flagged. The prepare_application_package tool gets you to "review and click submit" in one step instead.

Prerequisites

Python 3.10+
An MCP client — Claude Desktop is assumed below, but any MCP-compatible client works.
A Google account (for the Sheets tracker + Drive resume storage).
(Optional) free RapidAPI and/or Adzuna accounts for live job search.

Setup

Setup is: create a Google service account, share a Sheet + Drive folder + resume with it, (optionally) grab search-API keys, then point Claude Desktop at server.py. Budget ~10 minutes for the Google steps.

1. Google service account (~5 minutes)

Go to https://console.cloud.google.com → create a project (e.g. jobhunt).
Enable the Google Sheets API and Google Drive API (APIs & Services → Library).
APIs & Services → Credentials → Create credentials → Service account. Any name is fine; skip optional steps.
Open the service account → Keys → Add key → JSON. Save the file somewhere safe, e.g. ~/.config/jobhunt/service-account.json.
Copy the service account's email address (ends in .iam.gserviceaccount.com) — you'll share things with it next.

2. Google Sheet (the tracker database)

Create a blank Google Sheet named e.g. "Job Hunt Tracker".
Share it with the service account email as Editor.
Grab the spreadsheet ID from the URL: docs.google.com/spreadsheets/d/<THIS_PART>/edit.

The server creates the Jobs, Contacts, and Activity tabs with headers automatically on first use.

3. Google Drive folder + master resume

Create a Drive folder e.g. "Job Applications"; share with the service account as Editor. Copy its folder ID from the URL.
Put your master resume in Drive as a Google Doc (the everything-you've-ever-done version — tailoring trims it down). Share it with the service account. Copy its file ID.

4. Search API keys (optional)

JSearch (recommended — searches all boards at once): free at https://rapidapi.com → search "JSearch" → subscribe to the Basic (free) plan → copy your RapidAPI key into RAPIDAPI_KEY.

Adzuna: free at https://developer.adzuna.com → register → copy your App ID and App Key into ADZUNA_APP_ID and ADZUNA_APP_KEY.

Skip these if you'd rather paste postings in manually with jobhunt_add_job.

5. Install and configure

pip install -r requirements.txt

Copy claude_desktop_config.example.json into your Claude Desktop config (claude_desktop_config.json), then fill in the args path to server.py, the GOOGLE_SERVICE_ACCOUNT_FILE path, and the IDs/keys from the steps above.

Config file locations:

macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
Windows: %APPDATA%\Claude\claude_desktop_config.json (use "python" for command and double-backslash paths, e.g. C:\\Users\\you\\jobhunt-mcp\\server.py)
Linux: ~/.config/Claude/claude_desktop_config.json

Restart Claude Desktop (fully quit — on Windows/macOS check the tray/menu bar). The jobhunt tools appear once the server starts.

Environment variables

Everything is configured through the env block of the Claude Desktop config:

Variable	Required	Description
`GOOGLE_SERVICE_ACCOUNT_FILE`	✅	Absolute path to your service account JSON key
`JOBHUNT_SPREADSHEET_ID`	✅	ID of the tracker Google Sheet (shared with the service account)
`JOBHUNT_DRIVE_FOLDER_ID`	✅	ID of the Drive folder for saved documents
`JOBHUNT_MASTER_RESUME_FILE_ID`	✅	ID of your master resume Google Doc
`RAPIDAPI_KEY`	optional	RapidAPI key for JSearch (`jobhunt_search_all_boards`)
`ADZUNA_APP_ID`	optional	Adzuna App ID for `jobhunt_search_jobs`
`ADZUNA_APP_KEY`	optional	Adzuna App Key for `jobhunt_search_jobs`
`ADZUNA_COUNTRY`	optional	Adzuna country code (default `us`)

6. Verify it works

Test the server standalone with the MCP Inspector any time (no Claude Desktop needed):

npx @modelcontextprotocol/inspector python3 server.py

Then try jobhunt_get_master_resume (confirms Drive access) and jobhunt_list_jobs (confirms Sheets access). If both return without an auth error, you're set.

Tools

Tool	What it does
`jobhunt_search_all_boards`	Aggregated search across LinkedIn/Indeed/ZipRecruiter/Glassdoor via JSearch (RapidAPI)
`jobhunt_search_jobs`	Search live postings (Adzuna)
`jobhunt_add_job`	Save a posting to the tracker (paste the full description!)
`jobhunt_list_jobs` / `jobhunt_get_job`	Browse the pipeline / full record
`jobhunt_update_job_status`	Move through discovered → preparing → applied → interviewing → offer
`jobhunt_log_activity`	Log interviews, emails, notes; schedule follow-ups (`+7` = 7 days)
`jobhunt_get_followups_due`	Surface overdue and upcoming follow-ups
`jobhunt_add_contact` / `jobhunt_list_contacts`	Networking CRM per company
`jobhunt_get_master_resume`	Pull your master resume from Drive
`jobhunt_save_document`	Save tailored resumes/cover letters to Drive as Google Docs
`jobhunt_prepare_application_package`	Everything needed to tailor and apply, in one call

Example prompts to try

"Search all boards for SOC detection engineer roles near Dallas posted this week, and save the three best fits."
"Prepare an application package for job a1b2c3d4, tailor my resume to it, and save both docs."
"What follow-ups am I behind on?"
"I just finished the phone screen with Acme — log it and remind me to send a thank-you note tomorrow."

Security notes

The service account JSON key is a credential — keep it out of git and shared folders.
The service account can only touch files you explicitly shared with it.
Treat job descriptions fetched from the web as untrusted content; the server stores them as plain data.