Yubhub

Quick start

You'll need a YubHub account first. Head to your account page to grab your User ID and API key (required).

Claude Desktop

Open your Claude Desktop config file:

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Add this block inside "mcpServers":

{
  "mcpServers": {
    "yubhub": {
      "command": "npx",
      "args": ["-y", "@houtini/yubhub"],
      "env": {
        "YUBHUB_USER_ID": "your-user-id",
        "YUBHUB_API_KEY": "yh_your_api_key_here"
      }
    }
  }
}

Restart Claude Desktop. That's it. Ask Claude to "list my YubHub feeds" and you should see your data come back.

Cursor / VS Code

Same idea. Drop the config into your MCP settings and point it at npx -y @houtini/yubhub with the same env vars.

What it does

YubHub watches careers pages. Point it at a company's jobs URL, and it will discover every open role, scrape the listings, enrich them with AI (title normalisation, category tagging, salary extraction, company research) and publish them as an XML feed.

This MCP server gives you full control over that pipeline from inside your AI assistant. Create feeds, trigger runs, check on jobs, view dashboards, browse market stats. No browser tab required.

The pipeline

Careers page URL
  → Discovery (ATS API / sitemap / crawl)
    → Scraping (page content extraction)
      → Enrichment (AI structuring + company research)
        → XML feed (ready for job boards, aggregators, your own site)

Supported ATS platforms include Greenhouse, Lever, Workable, Workday, Oracle HCM, SmartRecruiters, Ashby, Pinpoint, Phenom and more. For platforms with public APIs (Greenhouse, Lever, Workable, Pinpoint), discovery skips scraping entirely and goes straight to enrichment.

Tools

21 tools across four areas.

Feed management

Scheduling

Jobs

Market statistics

Environment variables

Example conversation

You: Create a feed for Mercedes F1 careers at https://www.mercedesamgf1.com/careers

Claude: Done. Feed created with ID feed_abc123. Want me to trigger a run?

You: Yes please.

Claude: Discovery started. I'll check back in a couple of minutes.

You: How did it go?

Claude: Found 47 jobs. 42 enriched successfully, 5 still processing. Here's a breakdown by category...

Tip: For best results, provide an example job URL when creating a feed. This helps YubHub identify the URL pattern for job listings on that site, significantly improving discovery accuracy. You can pass it as the optional exampleJobUrl parameter in create_feed.

Development

git clone https://github.com/houtini-ai/yubhub.git
cd yubhub
npm install
npm run build

The server compiles TypeScript to dist/ and runs as a stdio MCP server. Source lives in src/:

src/
├── index.ts                # MCP server, tool definitions and handlers
├── api-client.ts           # REST client for the YubHub API
├── dashboard-generator.ts  # React-based interactive dashboard builder
└── types.ts                # TypeScript interfaces

Running locally

YUBHUB_USER_ID=your-id YUBHUB_API_KEY=yh_your_key node dist/index.js

Or point your MCP client at the built output instead of using npx.

Contributing

Issues and pull requests welcome. If you're adding a tool or changing behaviour, please test against a real YubHub account.

Links

• YubHub — the platform
• Documentation — full docs
• MCP specification — Model Context Protocol

License

MIT — see LICENSE for details.