PayIT

PayIT is a non-custodial Telegram wallet — hold USDC/EURC on Arc, spend in local currency, no app, no middleman. Send money, get paid, run your business, all in one chat

Comment

PayIT — Project Story

Inspiration

I'm based in Abuja, Nigeria, where I've watched the same problem play out for years: people earn in naira, watch its value erode month over month, and have no easy way to protect it. Small business owners around me still run invoicing and payroll out of WhatsApp messages and spreadsheets, manually chasing customers to confirm who paid what. I've spent time in Web3 infrastructure as an RC DAO Champion at NEAR Protocol, which showed me what non-custodial systems could offer ordinary people if the experience were simple enough. But almost nobody outside crypto-native circles can actually use stablecoins today — it requires an app download, an exchange account, and a tolerance for seed phrases. When I saw the Build with Gemini XPRIZE, it was the push to finally build the version of this idea I'd been sketching for a while: a wallet that lives inside an app people already have open all day — Telegram — and that understands people in plain language instead of forcing them to learn wallet mechanics.

What It Does

PayIT is a non-custodial Telegram wallet. Users hold USDC and EURC on Arc (Circle's stablecoin-native L1) while sending, receiving, and invoicing in their local currency. There's no separate app to install and no custodial exchange holding funds on their behalf. You talk to it like a person: "send 5000 naira to Maria," a voice note, or a photo of an invoice all work, because a multimodal AI agent — using the Gemini API alongside NVIDIA-hosted models for different steps of the pipeline — interprets the request and turns it into a structured on-chain payment. On the business side, PayIT includes invoicing with auto-reconciliation, payroll scheduling, and cash flow reporting, all inside one chat.

How I Built It

I started building PayIT the week the hackathon opened, as a solo developer, using the 90-day window to move fast on an idea I'd already thought through conceptually. The stack: Node.js with Telegraf for the Telegram bot layer, ethers.js for signing and broadcasting to Arc, and an AI agent layer that combines the Gemini API for [intent understanding / multimodal parsing] with NVIDIA-hosted models for other steps in the pipeline, chosen for speed and cost at this stage. Key design decisions:

Non-custodial by default. Each user's private key is encrypted client-side with a PIN-derived key (scrypt + AES-256-GCM). The server only ever stores an encrypted blob and salt, and keys are decrypted just-in-time, used once to sign, and destroyed immediately. Multimodal intent pipeline. A tiered system: fast heuristics for common commands, an LLM classification step for anything ambiguous, and a clarification flow that asks a targeted follow-up question rather than failing silently. Arc-native from day one. USDC as native gas removes the separate-gas-token friction that makes small, everyday transactions impractical on most chains.

Challenges I Faced

Building non-custodial infrastructure fast, without cutting the one corner that matters. Every shortcut that would make onboarding smoother — server-side key backup, simplified recovery — is a shortcut that reintroduces custody, which defeats the point. Holding that line while still shipping quickly took real discipline. Choosing the right model for the right step. Routing between Gemini and NVIDIA-hosted models for different agent functions meant thinking carefully about latency, cost, and reliability for each specific task rather than defaulting to one provider everywhere. Designing for real money with no revenue yet. PayIT is pre-revenue today. The near-term challenge isn't just shipping features, it's proving the ON/OFF-ramp flow can convert into real, monetizable transaction volume within the hackathon window.

What's Next

The immediate priority is closing the loop from wallet to real fiat: connecting a local payment partner for NGN off-ramp, so USDC balances can convert to real spendable naira, which is also the point where PayIT starts generating transaction-fee revenue. Alongside that, I'm expanding the SME invoicing flow, since that's the clearest path to recurring, monetizable usage from business users rather than one-off individual transfers.

Built With

Share this project:

Updates