Inspiration
We built MailHub to automate recruiting and reduce manual email work. The idea came from seeing recruiters spend hours on repetitive tasks—screening, scheduling, and follow-ups—that could be handled by AI. We wanted a system that evaluates candidates, coordinates interviews, and maintains a professional tone while freeing recruiters for higher-value work.
What it does
MailHub is an AI recruiter that: Evaluates candidates using Claude 3.5 on skills, experience, education, and location Categorizes emails into new applications, interview times, questions, and spam Coordinates interviews by forwarding availability to interviewers Generates responses using company information and conversation history Tracks candidate states through the hiring process with JSON persistence Sends emails via AgentMail API with error handling and fallbacks
How we built it
Backend: Python with AgentMail for email handling and Anthropic Claude 3.5 for AI Email processing: Polls AgentMail inboxes, categorizes with AI, and routes to handlers State management: JSON files for candidate data and processed email tracking AI integration: Custom prompts for evaluation, categorization, and response generation Error handling: Fallbacks for API failures and JSON parsing issues Testing: Local testing with MailHog and real email validation
Challenges we ran into
AgentMail API: Correct send method and parameter structure JSON parsing: Handling malformed AI responses with regex extraction Email categorization: Distinguishing new applications from interview times State persistence: Reliable candidate tracking across restarts Error handling: Graceful fallbacks for API failures and edge cases Accomplishments that we're proud of Full workflow: Application → evaluation → interview coordination → confirmation AI evaluation: Scoring candidates on multiple criteria with detailed reasoning Email automation: Sending and receiving emails with proper categorization State management: Persistent candidate tracking with JSON storage Error handling: Robust fallbacks and logging Real-time processing: Live email polling and processing
What we learned
AI integration: Prompt engineering and response parsing Email APIs: AgentMail structure and best practices State management: Persistence and recovery patterns Error handling: Fallbacks and logging strategies Workflow design: End-to-end automation patterns What's next for MailHub Calendar integration: Direct scheduling with Google Calendar/Outlook Advanced analytics: Hiring metrics and candidate pipeline insights Multi-company support: Multiple organizations and roles Mobile app: Recruiter dashboard and notifications AI improvements: Better evaluation and response generation Integration ecosystem: ATS and HR system connections