Andres BlancoAnalytics Engineering

Case Study

Job Seeking MCP

An AI-assisted job search workflow using a remote MCP connector and a Claude Skill.

The project helps make job searching more structured, repeatable, and targeted by connecting an AI assistant to approved job search sources and using the user's CV as context.

Download SkillView Install Guide

Problem

Job search needs better structure

The hard part is not only finding listings. It is translating a CV into a repeatable search strategy and reviewing results consistently.

Generic AI job searches are often too broad and hard to reproduce.

Users repeat the same prompts manually across sessions and tools.

Job matching is weak when role targets and keywords are not structured.

Candidates need a repeatable process that uses their CV as the source of truth.

Solution

A connector plus Skill workflow

The remote MCP owns job lead retrieval. The Claude Skill owns CV context, orchestration, ranking, and optional downstream outputs.

A Claude Skill reads the CV and extracts likely target roles and relevant search keywords.

The Skill sends structured search context to a remote Job Seeking MCP connector.

The MCP searches enabled job sources through approved public APIs and feeds.

Returned job leads are scored, ranked, and prepared for follow-up actions.

The workflow can optionally support a tailored CV version and Excel report.

Workflow

How it works

The process keeps the search context explicit: CV, roles, keywords, source results, scoring, then optional reporting outputs.

  1. 1Upload the latest CV.
  2. 2Extract target roles and relevant search keywords.
  3. 3Send structured search context to the Job Seeking MCP.
  4. 4Query enabled job sources through APIs and RSS feeds.
  5. 5Return matching remote job leads.
  6. 6Evaluate and score the results.
  7. 7Review a ranked list of opportunities.
  8. 8Optionally create a tailored CV draft and Excel report.

MCP capabilities

What the connector can do

The MCP is intentionally focused on retrieval, normalization, filtering, and diagnostics. It does not scrape job boards or automate applications.

  • List available job sources and enabled state.
  • Search public job postings from enabled sources.
  • Return structured compact job lead data.
  • Support Americas-compatible remote job search.
  • Provide compact diagnostics about source coverage and returned results.
  • Avoid scraping and use only allowed APIs and feeds.

Installation

How to use it with Claude Desktop

This walkthrough shows the current connector and Skill setup path. The MCP endpoint is public for the current version and does not require login.

  1. 1

    Install the Claude Desktop app.

  2. 2

    Log in to your account.

  3. 3

    Click Customize.

  4. 4

    Go to Connectors.

  5. 5

    Click Add Connector, then Add Custom Connector.

  6. 6

    Use Name: Job Seeking MCP and URL: https://job-seeking.andresblancog.com/mcp. No login is required for the current version.

    Claude custom connector setup screen for the Job Seeking MCP
    Add the remote MCP as a custom Claude connector.
  7. 7

    After adding it, the connector should appear as connected and show the available tools.

    Claude connector connected confirmation for the Job Seeking MCP
    Confirm the connector is connected and tools are available.
  8. 8

    Go to Skills.

  9. 9

    Click Add and upload the Job Search Assistant Skill.

    Claude Skill upload screen for the Job Search Assistant Skill
    Upload the Job Search Assistant Skill in Claude Desktop.
  10. 10

    Test the MCP connection with: "Use the Job Seeking MCP connector to list available sources."

  11. 11

    Then use the Skill with: "Can you use the job search assistant skill? I am uploading my latest CV."

The Skill file is available from this page as a Markdown Skill definition. Download the Job Search Assistant Skill.

Prompts

Example prompts

These prompts are intentionally practical. They describe the action, context, and output format instead of relying on vague search requests.

Use the Job Seeking MCP connector to list available sources.
Can you use the job search assistant skill? I am uploading my latest CV.
Search for remote data analyst roles that match my CV.
Create a ranked report of the best matching roles.
Create a tailored CV version for the top match.

Architecture

Technical structure

The system keeps retrieval separate from judgment. The MCP returns structured leads; the Skill uses CV context to decide what is relevant.

Remote MCP server endpoint exposed over Streamable HTTP.
Source adapters for public APIs and RSS feeds.
Normalized, structured job lead responses.
Claude Skill as the orchestration layer.
CV analysis and keyword extraction.
Scoring and ranking logic outside the MCP source layer.
Optional document and Excel report generation.

Limitations

Current limitations

This is an early working version.
The current source list is intentionally limited.
Job availability depends on the connected sources.
The project does not use web scraping.
Matching quality depends on CV quality and target role clarity.

Roadmap

Planned improvements

Add more public job sources and ATS board adapters.
Improve scoring logic and matching diagnostics.
Add better filters by region, seniority, salary, and tech stack.
Add authentication if needed for broader usage.
Improve Excel and CV reporting outputs.
Add stronger observability and source health diagnostics.

Tech stack

Built as a focused MCP workflow

The current project is a portfolio MVP: useful, constrained, and designed around allowed job sources rather than scraping.

MCPClaude DesktopClaude SkillsPythonStarletteUvicornhttpxPydanticYAML configurationAPIs/job feedsStructured dataExcel report generation