MCP Job-Tracker Server

Your job hunt as a conversation. This MCP server exposes your application tracker (a Google Sheet with 240+ applications) as tools any MCP client — Claude, Cursor, Windsurf — can call. Ask "which SF DevRel roles haven't replied in two weeks?" and get an answer, not a spreadsheet stare-down.

Why this exists

Spreadsheets don't answer questions. This turns a static tracker into an assistant for the hunt itself: reply triage, stale-lead detection, pipeline stats — all callable by the AI tools you already live in.

Related MCP server: jobwise

For AI builders (read this first)

Deliberately thin: the MCP SDK does the protocol, Google Sheets is the database, and each tool is one focused function. No framework, no ORM.

• Runtime: Node 20+, TypeScript, @modelcontextprotocol/sdk

• Transport: stdio (Claude Desktop / Cursor) — Streamable HTTP comes in v2

• Data: Google Sheets API reading the tracker tab; row 1 is the header

• Auth: GOOGLE_SERVICE_ACCOUNT_JSON env var (service account with viewer access to the sheet)

Repo structure

mcp-job-tracker/
├── src/
│   ├── index.ts      # MCP server entry: registers tools, starts stdio transport
│   ├── sheets.ts     # ← Sheets backend: fetch rows, parse into Application[]
│   └── tools.ts      # ← the three tools (pure functions over Application[])
└── .env.example

Tools

search_applications

Find applications by company, role, status, or channel.

{ "query": "DevRel", "status": "Sent", "limit": 10 }

reply_status

Which applications got replies? Correlates tracker rows with Gmail threads (by company name match).

{ "since": "2026-09-01" }

Returns: { company, role, status, replied: true|false, replyDate } per row.

stale_leads

Applications with no reply after N days (default 14), sorted oldest first — your follow-up list.

{ "days": 14 }

pipeline_stats (v2)

Counts by status/channel/week. "How many did I send last week? What's my reply rate?"

Application type

interface Application {
  date: string;       // 2026-10-04
  company: string;    // "Asana"
  role: string;       // "Developer Advocate"
  channel: "Portal" | "Email";
  status: "Submitted" | "Sent" | "Skipped" | "Blocked" | "In progress";
  comp: string;       // "$182–207K"
  notes: string;
}

Setup

npm install
cp .env.example .env   # add SHEET_ID + GOOGLE_SERVICE_ACCOUNT_JSON
npm run dev

Claude Desktop config:

{
  "mcpServers": {
    "job-tracker": { "command": "node", "args": ["/path/to/mcp-job-tracker/dist/index.js"] }
  }
}

Status: v1 built and tested

All four tools are implemented, typechecked, and covered by tests (npm test — 7/7 passing):

• search_applications — substring search across company/role/notes, optional status filter

• stale_leads — contacted apps older than N days (default 14), oldest first

• reply_status — Gmail correlation per company (best-effort name match, 90d lookback)

• pipeline_stats — totals by status/channel, contacted count, reply rate, weekly volume

Run the server: npm run dev (needs SHEET_ID + GOOGLE_SERVICE_ACCOUNT_JSON). Run the tests: npm test. Typecheck: npx tsc --noEmit.

Gmail note: reply_status needs Gmail read access. For a personal Gmail account, use OAuth2 user credentials — set GMAIL_OAUTH_TOKEN_JSON, GMAIL_OAUTH_CLIENT_ID, GMAIL_OAUTH_CLIENT_SECRET (see .env.example). Service accounts only work with Google Workspace domain-wide delegation.

Tech stack

TypeScript MCP SDK Google Sheets API Gmail API

Roadmap

• v1: stdio server + search_applications + stale_leads

• v2: reply_status via Gmail correlation

• v3: pipeline_stats + Streamable HTTP

• v4: "draft follow-up email" tool

License

MIT