New Hire Onboarding Assistant

Answers any new-hire question instantly: a production-ready multi-agent onboarding assistant that smartly routes queries to specialized LLMs, surfacing code, docs, policies, and teammates in seconds.

Comment

Inspiration

Broken onboarding costs companies $30 000+ per engineer and drains senior developers’ time. When Google released the Agent Development Kit, we realized a multi-agent AI could give new hires the contextual answers humans usually provide—without burning out the team. That spark became the New Hire Onboarding Assistant.

What it does

The New Hire Onboarding Assistant is a production-ready multi-agent AI system that transforms software engineer onboarding from weeks of confusion into days of confident productivity.

Core Capabilities: CodebaseNavigator: Explains repository architecture, analyzes dependencies, and teaches coding standards with real examples.

DocumentationAssistant: Instantly finds relevant wiki pages, API docs, and tutorials across all company knowledge bases.

TroubleshootingCopilot: Diagnoses errors, provides step-by-step solutions, and runs system diagnostics.

PolicyGuide: Handles HR policies, security guidelines, and compliance requirements (GDPR, SOC2, PCI-DSS).

TeamIntegrator: Connects new hires with the right teammates, schedules meetings, and provides team culture insights.

Ticket Management: Full CRUD operations on development tickets through integrated ToolBox API GitHub Integration: Repository search, issue tracking, and pull request management via MCP protocol.

Real Interactions: "How does authentication work in our system?" → Detailed walkthrough with actual code examples "I'm getting a database connection error" → Error analysis with specific troubleshooting steps "Who can help me with Python expertise?" → Team member recommendations with contact info and availability "Show me all critical priority tickets" → Live ticket data with filtering and analytics

How we built it

Architecture: Framework: Google Agent Development Kit (ADK) with 6-agent orchestration Database: PostgreSQL with JSON document storage containing 100+ realistic company documents Model: Gemini 2.5 Flash for all agents with specialized instructions

Technical Stack: Backend: Python with ADK framework, psycopg2 for database operations External Integrations: ToolBox API (16 ticket tools), GitHub MCP protocol, Google Search Data Layer: Structured JSON documents across 5 categories (teams, codebase, documentation, policies, troubleshooting) Security: Read-only API access, environment-based configuration, secure credential management

Database Design: Created comprehensive sample data including 13 detailed team member profiles, 6 repository descriptions with code snippets, complete HR policies, security guidelines, and troubleshooting solutions with realistic error patterns. Agent Orchestration: The root agent intelligently routes queries to specialized sub-agents while maintaining direct access to external tools. Each agent has domain-specific expertise with carefully crafted instructions that balance technical accuracy with human warmth.

https://github.com/Mahesh5h9/newhire_onboarding_assistant

Challenges we ran into

The MCP Integration Nightmare: The biggest challenge came three weeks before completion when StdioServerParameters imports suddenly broke across all MCP integrations. What had been working perfectly for calendar, GitHub, and PostgreSQL connections just stopped functioning. I spent three sleepless days trying every possible import combination while StreamableHTTPConnectionParams continued working fine. This forced a complete architectural pivot away from local MCP servers to HTTP-based integrations.

Other Technical Hurdles: Database Performance: Optimizing JSON queries across 100+ documents while maintaining search speed. Agent Personality: Crafting instructions that feel genuinely helpful rather than robotic. Production Security: Implementing proper access controls and credential management for real API integrations. Error Handling: Building robust fallbacks for external API failures.

Accomplishments that we're proud of

Production-Ready System: Built a fully functional multi-agent system with real database and API integrations. Created 100+ realistic company documents that feel authentic and comprehensive. Implemented 16 ToolBox tools for complete ticket management functionality. Achieved seamless agent orchestration with intelligent routing between specialists.

Technical Excellence: Zero hardcoded responses - all answers come from dynamic database queries. Real external integrations with GitHub, ToolBox API.

User Experience Innovation: Each agent has a distinct personality that feels genuinely helpful. Step-by-step guidance that builds confidence rather than overwhelming users. Proactive suggestions for related resources and next steps.

What we learned

Technical Insights: ADK's agent orchestration is incredibly powerful but requires careful tool distribution strategy. PostgreSQL JSON operations can be blazingly fast with proper indexing and query optimization. Dependency management is critical - always have fallback strategies for external integrations. MCP protocol has immense potential but needs more stable implementation paths.

Product Realizations: Multi-agent systems mirror how humans naturally seek help from different specialists. Real data beats perfect algorithms - authentic company information makes the experience believable. Context matters more than accuracy - understanding user intent enables better responses. Security can't be an afterthought - every tool needs proper access controls from day one.

Human-Centered Design: New hires don't just need information—they need confidence and encouragement. The best onboarding feels like guided discovery, not information dumping. Personality in AI requires weeks of iteration to feel natural rather than scripted.

What's next for New Hire Onboarding Assistant

Fix MCP StdioServerParameters integration for richer local tool connectivity. Advanced analytics dashboard for tracking onboarding progress and identifying knowledge gaps. Slack/Teams integration for seamless workflow embedding.

Advanced Features: Learning path generation that creates personalized curricula based on role and experience Peer matching system that connects new hires with similar backgrounds or interests. Progress tracking with milestone celebrations and achievement systems.

Enterprise Scaling: Custom data connectors for various HR systems, wikis, and knowledge bases. Advanced compliance modules for industry-specific requirements (healthcare, finance). Integration marketplace for popular development tools and platforms.

Built With

Share this project:

Updates

Submission history