About this project
The problem: India's NBFCs lose qualified borrowers to slow, opaque, consent-blind loan origination. KYC, fraud screening, credit-bureau checks, bank-statement analysis, affordability math, and underwriting are stitched together by call-centre agents and brittle scripts—with DPDP consent treated as a checkbox, not an enforced control. Vitta is an MCP-native NBFC loan-origination server built on NitroStack and deployed to NitroCloud. It exposes the "hi → signed sanction letter" journey as MCP primitives, letting any client (NitroChat, Claude, ChatGPT Apps) act as the reasoning agent while Vitta is the deterministic capability layer—qualify_lead, record_consent, verify_kyc, screen_fraud, pull_bureau, fetch_bank_statements, compute_affordability, underwrite, generate_offers, create_sanction_letter, and audit, orchestrated tool by tool. Vitta ships 16 Tools, 5 Resources, and 5 Prompts localized in English, Hindi, and Malayalam. Its engine is a pure deterministic scorecard—FOIR-based underwriting, reducing-balance EMI, hard-negative overrides—using rules, not black-box ML, so every decision is explainable via reason codes, alongside live ECB/Frankfurter FX rates and redacted, append-only audit trails. The consent gate is the core innovation: pull_bureau and fetch_bank_statements refuse to run without a scoped, time-boxed, HMAC-signed token, making DPDP compliance executable in code. A What-If simulator re-runs underwriting on hypothetical changes—like closing a ₹25,000 EMI—without touching the real case. Every tool is a drop-in seam for real bureau, Account Aggregator, and e-sign APIs. Being MCP-native, one server serves many AI assistants at once—delivering faster, auditable, consent-first origination for NBFCs and provable compliance for regulators.
BFSI & FinTech track
Build AI solutions for banking, payments, insurance, fraud detection, lending, and financial inclusion.