IRCTC Ticket Booking AI Agent

I’m building an end-to-end conversational agent that can take a traveller from the moment they say “I need a ticket” all the way to receiving the confirmed IRCTC PNR—without manual intervention from my side once the user approves the purchase. Scope of work The agent must live simultaneously on my website chat widget, on WhatsApp, and inside an existing mobile app. Through natural conversation it should collect travel date, route, train preference, passenger and berth details, plus the user’s preferred payment method and contact information. After gathering those details it should log in to IRCTC, search availability, select the requested train or the closest alternative, and complete the payment flow. Finally it returns the PNR and a receipt back to the same chat thread. Core requirements • Multichannel deployment: website, WhatsApp, and mobile SDK. • Secure storage of user credentials and payment tokens; PCI-DSS compliant flow for card or UPI. • Resilient handling of IRCTC captcha, session expiry, and peak-hour throttling. • Conversation engine can be rule-based, LLM-powered, or a hybrid, as long as the hand-off between data collection and booking API is seamless. • Dashboard or logs for me to monitor each step, with retry logic and error alerts. Deliverables 1. Source code for the conversational front-ends and the booking microservice. 2. Deployment scripts (Docker/Kubernetes preferred) and minimal cloud architecture diagram. 3. Test suite proving a full booking cycle succeeds on the live IRCTC portal. 4. Brief hand-over session and documentation. Acceptance criteria • A test user can request a ticket via any of the three channels and receive a confirmed PNR in under two minutes. • System gracefully prompts for missing data and reports meaningful errors from IRCTC. • All collected personal and payment data is encrypted at rest and in transit. If you have previous experience bypassing IRCTC captcha or integrating with their unofficial APIs, please mention it—this will speed up short-listing.