2026 Flagship ยท Autonomous Job Discovery Agent

FreeJobAgent

Autonomous job discovery and semantic matching agent with Model Context Protocol (FastMCP) and Gemini API.

Year: 2026 | Role: Creator and Lead Engineer. Architected the agent workflow, implemented the MCP server, and built the extraction and evaluation pipeline.

View source on GitHub

Project Overview

FreeJobAgent automates technical job discovery. It scrapes live job postings, evaluates candidate resume fit using semantic similarity and LLM reasoning, and drafts customized application pitches.

Problem Statement

Job seekers spend hours browsing fragmented job boards, manually reading qualification criteria, and tailoring applications without objective match scoring.

Solution Architecture

Engineered an autonomous agentic pipeline combining Apify LinkedIn job extraction, Model Context Protocol (FastMCP) for decoupled tool context retrieval, and LangGraph ReAct loops for multi-step profile evaluation.

  • Candidate Parser: Extracts skills, experience, and criteria vectors from the applicant resume.
  • Job Extraction Pipeline: Interfaces with Apify actors to pull live, localized job listings.
  • FastMCP Server: Exposes job parsing, filtering, and scoring tools cleanly to the Gemini API.
  • ReAct Evaluation Loop: Evaluates semantic alignment, skill overlaps, and gaps to generate match confidence percentages.
  • Streamlit Dashboard: Provides interactive filtering, criteria tuning, and application draft exports.

Tech Stack & Engineering

  • LangGraph
  • LangChain
  • Model Context Protocol (FastMCP)
  • Google Gemini API
  • Apify
  • Streamlit
  • Python

Key Implementation Features

  • Automated LinkedIn job extraction and structured parsing via Apify.
  • Standardized Model Context Protocol tool interface with FastMCP.
  • Multi-dimensional candidate-to-job semantic fit scoring.
  • Streamlit dashboard with customizable role, location, and confidence thresholds.

Technical Challenges & Solutions

  • Handling unstructured job description formats: Implemented schema validation with Pydantic to normalize disparate recruiter formats before evaluation.
  • Avoiding context window saturation during multi-job batch analysis: Designed a multi-stage filter that screens keywords before passing high-probability matches to the LLM.

Verified Results

  • Successfully extracted and ranked live technical job postings in automated validation runs.
  • Demonstrated practical tool decoupling using the emerging Model Context Protocol standard.