Skip to main content
Glama
JobsDirect MCP

JobsDirect MCP

by weblinkify

JobsDirect MCP

Live demo: https://jobs-direct-mcp.vercel.app (mock mode, runs fully in the browser) · deploys automatically on every push to main

An MCP server that lets Claude Desktop, Cursor or any MCP client search six job portals, score how well you fit each posting, and prepare the application: a tailored resume, a cover letter, a Word file and a tracker entry. It runs locally; your CV never leaves your machine.

MCP server

Python 3.10+ · FastMCP · httpx · BeautifulSoup · python-docx

Web demo

React 18 · TypeScript · Tailwind CSS · Vite, deployed on Vercel

Portals

LinkedIn · Himalayas · Remote.co · CareerHound · HiredChina · Kariyer.net


Tools

Tool

What it does

search_jobs

Searches one or all portals. Filters by workplace (remote/hybrid/onsite) and posting age, and drops postings that say they are closed.

get_job_details

Full description, criteria and open/closed status for a LinkedIn job ID or any posting URL.

evaluate_job_fit

4-tier ATS score (0-100) with a verdict and the reasons behind it.

tailor_resume

Re-ranks your own CV bullets by relevance, lists matched and missing keywords, flags bullets without a metric. It does not invent experience.

generate_cover_letter

Short letter citing your two most relevant achievements.

export_to_docx

Saves Markdown as a Word document.

prepare_job_application

Fetch + score + resume + letter in one call.

track_application

Add, update and list applications in job_applications.csv.

lookup_salary_benchmark

Indicative junior/mid/senior bands for US, EU, UK, Türkiye and remote.

The 4-tier fit score

Fit (0-100) =
  Tier 1  Industry affinity        40   your industry keywords vs. the employer's
  Tier 2  Qualifications gate      20   degree level + years of experience (a "3-5 years" range counts as 3)
  Tier 3  Skills overlap           30   tools and methods the posting names that your CV also names
  Tier 4  Language & eligibility   10   a missing required language fails the gate; clearance / citizenship
                                        wording asks you to confirm work authorisation

Verdicts: STRONG_MATCH (80+), MODERATE_MATCH (60+), WEAK_MATCH, plus gate verdicts that override the number: CLOSED (closure notice or a past deadline, in English or Turkish), LANGUAGE_GATE_FAILED, VERIFY_ELIGIBILITY, NOT_ELIGIBLE and POOR_DOMAIN_FIT. A posting for crypto trading will not score well for a heavy-machinery project manager just because both mention "Agile".


Related MCP server: job-tracker

Install

git clone https://github.com/weblinkify/jobs-direct-mcp
cd jobs-direct-mcp
python -m venv .venv && source .venv/bin/activate     # Windows: .venv\Scripts\activate
pip install -e .

Add it to your MCP client (claude_desktop_config.json, Cursor mcp.json, …):

{
  "mcpServers": {
    "jobsdirect": {
      "command": "/ABSOLUTE/PATH/jobs-direct-mcp/.venv/bin/jobsdirect",
      "env": { "JOBSDIRECT_OUTPUT_DIR": "~/Desktop" }
    }
  }
}

On Windows the command is C:\\path\\jobs-direct-mcp\\.venv\\Scripts\\jobsdirect.exe. Files (Word exports, the tracker CSV) go to JOBSDIRECT_OUTPUT_DIR, else your Desktop, else ~/JobsDirect.

Example prompt

I'm looking for remote senior frontend roles in Europe. My CV is attached.
1. search_jobs for "senior frontend react" on all portals, last 7 days.
2. evaluate_job_fit for each result against my CV and pick the top 2.
3. tailor_resume and generate_cover_letter for the best one, then export_to_docx.
4. track_application with status "saved".

Web demo

cd frontend
npm install
npm run dev          # http://localhost:5173

The demo uses sample postings (fictional companies) but the scoring and writing are real: the browser runs a TypeScript port of the Python engine. Edit the profile and every posting re-scores. Tracked applications stay in your browser's local storage and export to CSV.

The two engines stay identical: fixtures/generate.py writes fixtures/ats_cases.json from the Python code, and both tests/test_parity.py and frontend/src/__tests__/parity.test.ts assert against it.

Deploy on Vercel

Import the repo in Vercel and set Root Directory to frontend. frontend/vercel.json handles the rest (Vite build, SPA rewrite, cache and security headers). No environment variables are needed.


Development

make install   # dependencies and git hooks
make check     # lint, types, tests and build: everything CI runs before merge
make audit     # pip-audit and npm audit
python fixtures/generate.py      # after changing jobsdirect/ats.py

Notes

  • Portal markup changes from time to time. Each portal fails on its own and is reported in portal_errors, so one broken source does not stop a search.

  • Respect each site's terms of use and rate limits; this tool fetches only public pages, one request per search.

  • Salary bands are indicative planning numbers, not live market data.

License

MIT

Available Tools

9 tools
evaluate_job_fitEvaluate Job FitB

Score candidate-to-job fit 0-100 with the 4-tier ATS engine.

Tiers: industry/domain affinity (40), degree + years gate (20), skills overlap (30), language and work-authorisation gate (10, a missing required language fails the gate).

Inputs: job_description (required), candidate_profile (required), work_authorized, candidate_languages. Returns: result

export_to_docxExport To DocxB

Save Markdown text (resume, cover letter, report) as a Word .docx in the output folder.

Inputs: title (required), content (required), filename. Returns: result

generate_cover_letterGenerate Cover LetterC

Draft a short cover letter that cites the candidate's two most relevant achievements.

Inputs: job_title (required), company_name (required), job_description (required), candidate_profile (required), +2 more. Returns: result

get_job_detailsGet Job DetailsA

Full description, criteria and open/closed status for a job ID (LinkedIn) or any posting URL.

Inputs: job_id_or_url (required). Returns: result

lookup_salary_benchmarkLookup Salary BenchmarkB

Indicative junior/mid/senior annual salary bands for a role in US, EU, UK, Türkiye or remote.

Inputs: role (required), location. Returns: result

prepare_job_applicationPrepare Job ApplicationB

One call: fetch the posting, score fit, tailor the resume and draft the cover letter.

Inputs: job_id_or_url (required), candidate_profile (required), candidate_name. Returns: result

search_jobsSearch JobsC

Search open job postings.

Inputs: query (required), limit, portal, location, +2 more. Returns: result

tailor_resumeTailor ResumeB

Re-rank the candidate's own CV bullets by relevance to the job, list matched/missing keywords, and flag bullets that need a metric. Returns Markdown (feed it to export_to_docx).

Inputs: candidate_cv (required), job_description (required), target_role, target_company. Returns: result

track_applicationTrack ApplicationC

Manage the applications CSV (job_applications.csv in the output folder).

Inputs: role, notes, action, job_id, +4 more. Returns: result

Tool Schema Changelog

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

Loading tool schema changelog…

    TDQS

    B3.2/5.0

    Scored across 9 tools

    Disambiguation3/5

    Most tools have distinct purposes, but prepare_job_application overlaps heavily with get_job_details, evaluate_job_fit, tailor_resume, and generate_cover_letter by bundling them into one workflow shortcut. Agents may be unsure whether to call the composite tool or the granular tools for a given request, though the descriptions make the tradeoff reasonably clear.

    Naming Consistency5/5

    All tool names use snake_case and follow a predictable verb_noun or verb_noun_phrase pattern, such as search_jobs, get_job_details, evaluate_job_fit, and generate_cover_letter. export_to_docx is a minor phrasing variation but still fits the consistent imperative naming style.

    Tool Count5/5

    Nine tools are well-scoped for a job-search and application-preparation server, with each major workflow step represented. The count is neither thin nor sprawling, and the composite tool does not inflate the surface excessively.

    Completeness4/5

    The surface covers the core lifecycle: search jobs, inspect details, evaluate fit, tailor resume, generate cover letter, export documents, track applications, and benchmark salary. Minor gaps remain around richer application tracking operations and explicit candidate-profile management, but agents can work around them.

    Maintenance

    ActivityMaintained
    ResponsivenessNo recent community issues

    Related MCP Connectors

    Related MCP Servers

    • F
      license
      Not graded
      quality
      D
      maintenance
      Enables Claude to parse CVs, search job boards (Remotive, Arbeitnow, Adzuna, Greenhouse/Lever), tailor resumes and cover letters, and prepare application packages with direct apply links—without ever auto-submitting. It runs 100% locally and free, storing jobs and applications as JSON files.
      -
    • A
      license
      A
      quality
      B
      maintenance
      Enables managing a job search through natural language: tracking applications, discovery leads, interview prep, and resume generation. Connects to Claude via MCP to read and update local Excel files and documents.
      35
      ISC
    • F
      license
      Not graded
      quality
      D
      maintenance
      Enables Claude Desktop to manage a job search end-to-end: find and score job listings, tailor resumes, generate application messages, and track application history, while leaving final external actions to the user.
      -
    • A
      license
      A
      quality
      C
      maintenance
      Enables parsing a master resume, analyzing job descriptions for ATS/match scores, generating truthful tailored resumes, tracking applications, and producing PDF reports through Claude Desktop.
      17
      MIT