convez.ai
MCP ServerProdutividadeIntermediário
benskamps

Career Compass MCP

benskamps/career-compass-mcp

Assistente IA de carreira local-first: currículo, pipeline e prep de entrevista

Ver no GitHub
InicialPluginScore · de 0 a 100

Servidor MCP que transforma Claude em um assistente completo de busca de emprego. Gerencia histórico de carreira em YAML local, gera currículos personalizados, rastreia aplicações e prepara para entrevistas — tudo com dados privados na sua máquina.

Instalação

$npx convez-ai benskamps/career-compass-mcp

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
7
Forks
0
Acompanhando
1
Saúde
42%
Último commit
há 3 dias
Versão
v2.15.1
Issues abertas
1
Contribuidores
1

Casos de uso

  • Preparação para painéis e entrevistas com análise de seu histórico
  • Geração automática de currículos e cover letters customizados por vaga
  • Rastreamento e análise de pipeline de aplicações com recomendações

Veredito

MCP bem arquitetado para candidatos seriamente envolvidos com busca de emprego. A abordagem local-first é um diferencial genuine de privacidade; o valor real emerge quando integrado com Claude para análise contextual de histórico, roles e gaps específicos.

README

A tradução para o português deste README ainda está na fila.

Career Compass MCP

License: MIT Node MCP

Your AI-native career co-pilot, built for Claude.

Career Compass turns Claude into a full job-search partner — one that knows your entire career history, tailors every resume and cover letter, tracks every application, and preps you for every interview. Initiate via conversation, onboard, and track multiple applications in moments.


Your data stays on your machine

Career Compass is local-first. Your career history lives in plain YAML files on your own disk, under the directory you point CAREER_DATA_PATH at (default ~/.career-compass). There is no account, no cloud sync, and no telemetry — nothing phones home.

Your real data is never committed to git: the .gitignore excludes data/career/ and data/pipeline/, so the only career data in this repo is the fictional sample under data/example/ (meet Alex Rivera). Read it, edit it, delete it — it's all just files you own.


What it feels like

You: I have a panel interview at Veridian Health on Friday — Director of Operations role.
     Can you prep me?

Claude: On it. Reading your career history now...

     [Generates 90-second pitch, 8 STAR stories matched to likely panel questions,
      company research brief, 10 questions to ask them, and a list of watch-outs
      based on gaps in your background — all in one response]
You: Here's a job posting I just found. [pastes posting]
     How well do I fit?

Claude: Fit score: 8.1/10. Here's why — and here's what they'll probe you on...

     [Returns matched strengths, honest gap analysis, talking points in their language,
      and a "day in the life" of what the role actually looks like]
You: Show me what needs attention in my pipeline today.

Claude: 3 things:
     - Meridian Logistics follow-up is overdue (8 days since you applied, referral from Marcus Chen)
     - Veridian panel is Friday — prep above
     - Novare rejection arrived — want me to draft a keep-the-door-open response?

Architecture

┌─────────────┐         ┌──────────────────────────────────────┐
│             │  MCP     │         MCP Server (Node.js)         │
│   Claude    │◄────────►│                                      │
│             │  stdio   │  Tools ─── Resume, Pipeline, Interview│
│             │         │  Resources ─ Career KB, Pipeline data │
└─────────────┘         │  Prompts ── Power-user shortcuts      │
                        │                                      │
                        │         ┌──────────────┐             │
                        │         │  File Store   │             │
                        │         │  (YAML files) │             │
                        │         └──────┬───────┘             │
                        └────────────────┼─────────────────────┘
                                         │ reads
                        ┌────────────────┼─────────────────────┐
                        │   Dashboard    │    (Next.js)         │
                        │                ▼                      │
                        │  Pipeline Kanban · Career KB Overview │
                        │  Analytics · Onboarding Wizard        │
                        └──────────────────────────────────────┘

Your data stays on your machine

Career Compass is local-first by design. Your real career data — résumé history, the companies you're talking to, salary numbers, interview notes — lives in plain YAML files on your own disk and never leaves it.

  • Where your data lives: ~/.career-compass/ by default (override with the CAREER_DATA_PATH environment variable). This directory is created and read on your machine. It is never uploaded anywhere, and it is not part of this npm package.
  • What ships in the package: only the MCP server code and a small set of fictional example files (data/example/ — the "Alex Rivera" persona). There is no real career data in the published package, and none can be — the package and your data live in different places.
  • The dashboard reads your data at request time, locally. The Next.js routes are force-dynamic, so the dashboard server reads your YAML when you open a page, from a server running on your own localhost. Your data is never baked into a build, never prerendered into static HTML, and never sent over the network. (There's a regression test, standalone-dynamic.test.ts, that guards exactly this.)
  • No telemetry. Career Compass makes no analytics or phone-home calls. The only thing that ever sees your career data is the Claude conversation you start, through the MCP server you run.

In short: this is a tool you run, on your machine, over your files. Treat ~/.career-compass/ like any private notebook — back it up, don't commit it to a public repo, and it stays yours.


Dashboard

v2.0 ships a local web dashboard — a visual layer on top of your Career KB and pipeline data. Open it alongside Claude or use it standalone to review and manage your search.

Pipeline Kanban Application Detail Career KB Analytics

Four views:

Pipeline kanban board — All your applications laid out by stage (Exploring → Applied → Screening → Interviewing → Offer → Closed). Drag to advance stages. See stale applications at a glance. Click any card to drill in.

Application detail view — Full timeline for a single application: every status change, contacts associated with the role, notes from conversations, and next action. Everything Claude knows about that opportunity in one place.

Career KB overview — Visual summary of your career data: skills radar (proficiency × recency), experience timeline, testimonials, and a completeness indicator showing what Claude has to work with. Useful for spotting gaps before you onboard a new role.

Analytics — Funnel conversion rates, response rates by source, time-in-stage averages, and source effectiveness. See which channels are actually working.

Open the dashboard:

npm run dashboard

Quick Start

Status: install from source. Career Compass is not published to npm yet, so there is no npm install -g path today. You build it from this repo with a couple of commands. When the package lands on npm, this section will switch to a one-line global install.

1. Build from source

git clone https://github.com/benskamps/career-compass-mcp.git
cd career-compass-mcp
npm install
npm run build

npm run build compiles the MCP server (TypeScript → build/) and the Next.js dashboard. See Building from Source for the partial-build options.

2. Try it now (no setup, ~1 minute)

Want to see the dashboard before wiring up Claude? Point it at the bundled sample data and open it:

CAREER_DATA_PATH=data/example npm run dev:dashboard

That loads the full dashboard — pipeline kanban, Career KB overview, analytics — populated with the fictional Alex Rivera sample, so you can click through every view in under a minute before adding any of your own data.

3. Add to Claude

Add this to your ~/.claude.json (Claude Code) or equivalent MCP config, pointing command at the built server:

{
  "mcpServers": {
    "career-compass": {
      "command": "node",
      "args": ["/path/to/career-compass-mcp/build/src/index.js"],
      "env": {
        "CAREER_DATA_PATH": "/Users/you/career-data"
      }
    }
  }
}

Point CAREER_DATA_PATH anywhere you like — Career Compass will initialize the directory on first run. Leave it unset to use the default, ~/.career-compass.

4. Open the dashboard

npm run dashboard

This opens the dashboard in your browser at http://localhost:3141 (or the next available port). Use it to finish onboarding and get a visual overview of your pipeline.

5. Onboard (first conversation)

Open Claude and say:

"Set up my Career KB. Here's my resume:" [paste your resume]

Claude will:

  • Extract your work history, achievements, and skills into structured YAML
  • Ask clarifying questions about gaps or unclear metrics
  • Save everything to your CAREER_DATA_PATH

Then open the dashboard to complete the second phase: the onboarding wizard walks you through setting job targets, salary expectations, and confirming your skills inventory. Claude pulls the raw data from your resume; the wizard fills the gaps that a resume doesn't naturally contain.

That's it. From there, every tool has full context on who you are.


Data structure

Career Compass stores all data as YAML files under CAREER_DATA_PATH (default ~/.career-compass):

~/.career-compass/
├── career/
│   ├── profile.yaml        # who you are, what you're targeting
│   ├── experience.yaml     # roles, achievements (metrics + context + impact)
│   ├── skills.yaml         # skills with proficiency and recency
│   ├── education.yaml      # degrees, certifications, coursework
│   ├── projects.yaml       # portfolio projects
│   └── testimonials.yaml   # quotes and recommendations
└── pipeline/
    └── applications.yaml   # all job applications

This is your single source of truth — built once, enriched over time, read by every tool. You never need to edit these files by hand: use ingest_document to add data by pasting documents, or just ask Claude to update specific sections.

See data/example/ in this repo for a fully populated sample (the fictional Alex Rivera).


Dashboard CLI

Run via the npm script (or node build/bin/cli.js dashboard directly):

npm run dashboard                                 # open dashboard (default port 3141)
node build/bin/cli.js dashboard --port 3000       # specify port
node build/bin/cli.js dashboard --no-open         # start server without opening browser

Tools

ToolWhat it does
explore_opportunityAnalyzes a job posting against your Career KB — fit score, matched strengths, gaps, talking points, day-in-the-life, red flags
research_companyBuilds an intelligence brief: product, culture, funding, interview process, strategic fit
tailor_resumeGenerates an ATS-optimized, tailored resume from your KB — standard, federal, academic, or functional formats
generate_cover_letterWrites a personalized cover letter with your actual achievements woven in
format_for_atsReformats resume content for specific ATS systems: Workday, Greenhouse, Lever, LinkedIn, iCIMS, Taleo
manage_pipelineTracks applications from discovery through offer — add, update, list, stats, next actions
classify_emailClassifies a job-search email and extracts contacts, dates, and suggested pipeline updates
prepare_interviewFull interview prep: opening pitch, STAR stories, likely questions, company alignment, questions to ask
evaluate_offerBreaks down total comp, compares to market, builds negotiation strategy, drafts counter scripts
ingest_documentExtracts achievements from any document: performance review, award email, LinkedIn recommendation, project summary
generate_rejection_responseDrafts a graceful response that keeps the door open and maintains the relationship

Resources

Claude can read these directly (e.g., "read my career profile"):

ResourceURIContents
Career Profilecareer://profileName, contact, summary, targets, preferences
Work Experiencecareer://experienceFull history with achievements
Skills Inventorycareer://skillsSkills with proficiency and recency
Projectscareer://projectsPortfolio with outcomes
Educationcareer://educationDegrees, certifications, coursework
Testimonialscareer://testimonialsQuotes, recommendations
Full KBcareer://fullEverything above in one read
Pipelinecareer://pipelineAll applications with status

Prompts

Power-user shortcuts (appear in Claude's prompt menu):

PromptWhat it does
resume-tailorDrop in a posting → get a tailored resume
interview-coachCompany + role + interview type → full prep package
negotiation-coachPaste an offer → get analysis, strategy, and counter scripts

Configuration

Env varDefaultDescription
CAREER_DATA_PATH~/.career-compassDirectory where your career and pipeline YAML files are stored

Building from Source

git clone https://github.com/benskamps/career-compass-mcp
cd career-compass-mcp
npm install
npm run build

npm run build compiles both the MCP server (TypeScript → build/) and the Next.js dashboard (dashboard/.next/). To work on just the MCP server, use npm run build:mcp. For dashboard development with hot reload, use npm run dev:dashboard.

Then point your MCP config to node /path/to/career-compass-mcp/build/src/index.js.

Use npm run dev during development — TypeScript watch mode recompiles on save. Use npm run inspect to open the MCP Inspector (web UI for testing tools interactively).


Why Career Compass

Job searching is one of the highest-stakes, most document-intensive activities most people do — and most tools treat it as a data entry problem. Spreadsheets for tracking. Templates for resumes. Generic advice for interviews.

Career Compass treats it as a knowledge problem. Your career history is a corpus. Every application is a retrieval and synthesis task. Every interview is a pattern-matching problem against a known dataset (the posting) and a known corpus (your KB).

The Career KB is your single source of truth — built once, enriched over time, and used by every tool. A tailored resume draws from it. Interview prep draws from it. Cover letters draw from it. The pipeline tracks against it. Nothing gets lost because it's never in a tab you'll close.


Development

git clone https://github.com/benskamps/career-compass-mcp.git
cd career-compass-mcp
npm install
cd dashboard && npm install && cd ..

# MCP server (TypeScript watch)
npm run dev

# Dashboard (Next.js + Turbopack)
npm run dev:dashboard

# MCP Inspector (test tools interactively)
npm run inspect

# Tests
npm run test:mcp        # MCP server tests
npm run test            # Dashboard tests

# Storybook (component library)
npm run storybook

# Use example data for development
CAREER_DATA_PATH=data/example npm run dev:dashboard

Contributing

Issues and PRs welcome. If you add a new tool, register it in src/server.ts and follow the pattern in any existing tool file — each tool returns a structured prompt that Claude acts on with the full KB in context.


License

MIT


Part of the Brokenbranch Lab — Ben Schippers' workshop of AI-native tools and research.

Ver o original no GitHub

Bom para

Relacionados