Retriis — Autonomous Agentic Post-Purchase Manager

Tagline:
"Returns that feel like magic — autonomous post-purchase for P2P marketplaces."

Inspiration

Peer-to-peer marketplaces are booming across Nigeria and Africa, but the post-purchase experience is still broken. Buyers face frustrating forms and slow support, while sellers drown in manual disputes and refund requests.

After seeing countless stories like Dave (a buyer in Lagos whose phone case cracked after one week) struggling with poor service and Malcolm (the seller) losing hours to manual ticket handling, I knew there had to be a better way.

Retriis was born from the need to make returns and exchanges feel as smooth as chatting with a super-smart store manager — fully autonomous, policy-driven, and built for the African P2P reality.

What it does

Retriis is a fully agentic post-purchase engine for P2P marketplaces. Buyers chat naturally (“My phone case cracked after one week”), upload photos, and an autonomous AI agent instantly:

Verifies the order
Applies seller’s plain-English policies
Processes refunds via Paystack or Stripe
Generates shipping labels
Issues loyalty store credit
Coordinates follow-ups and updates analytics

It supports both an embeddable widget (one JS snippet for any marketplace) and a built-in lightweight P2P storefront. Sellers manage policies via natural language chat, while marketplace operators get powerful analytics and dispute mediation.

How we built it

I built Retriis entirely using MeDo — the AI-powered app builder.

I started with a detailed 15k+ byte requirements document and a comprehensive starter prompt. Through structured multi-turn conversations, I iteratively built:

The autonomous AI agent core
Multi-turn conversational UX with photo upload
Seller policy configuration chat
Real-time analytics dashboard with pattern detection
Embeddable widget + full P2P storefront

Best part MeDo generated: The seamless embeddable widget accessible from the Settings page — operators simply copy one JavaScript snippet and paste it on any site.

Plugins & Integrations:

MeDo native Paystack and Stripe plugins
MeDo Email and Calendar plugins
Supabase for authentication and role-based access

Challenges we ran into

Orchestrating complex autonomous agent workflows (order verification → policy check → payment → label generation → loyalty) while keeping it reliable.
Designing smooth multi-turn conversations that feel natural yet handle edge cases.
Balancing a great buyer experience with powerful seller controls and analytics.
Ensuring proper escrow release logic and high-value item escalations.

Accomplishments that we're proud of

A fully functional autonomous return cycle that completes in under 60 seconds.
Seamless embeddable widget that works on any marketplace.
Beautiful, intuitive UI with real-time analytics and pattern detection.
Complete end-to-end flows using real payment plugins (Paystack + Stripe).
Realistic demo data and role-based authentication (Buyer, Seller, Operator).

What we learned

The power of natural language development with MeDo — you can build production-grade apps incredibly fast.
How critical autonomous agents are for reducing operational friction in marketplaces.
The importance of Nigeria-first solutions (Paystack priority, local currency, Lagos-focused flows).