Encore

Redefining the landscape of 2ndhand e-commerce, Encore's agent Cora recommends, drafts, markets, & negotiates your secondhand sales so you step in only where it matters. Good things deserve a 2nd life

Comment

💫 Inspiration

Selling secondhand goods benefits everyone: sellers recover value, buyers get affordable products, and useful items avoid becoming waste. Yet listing an item often means writing descriptions, answering repetitive questions, handling offers, and coordinating buyers.

Encore started with one question: What if selling an item were as easy as taking a picture?

🚀What it does

Encore is a secondhand marketplace powered by Cora, an AI selling assistant.

Sellers photograph an item, and Gemini creates an editable listing draft. They confirm the details, set an asking price and private minimum, add pickup preferences, and publish.

Cora then:

  • Answers questions using seller-confirmed facts.
  • Negotiates within the seller’s private boundaries.
  • Escalates anything uncertain instead of guessing.
  • Sends acceptable offers to the seller for approval.
  • Never reveals private instructions or completes a sale independently.

Encore also includes marketplace filters, personalized recommendations, inventory tracking, seller insights, notifications, quantity-aware reservations, and optional Solana devnet payments.

🛠️ How I built it

  • Frontend: React, TypeScript, Vite, and React Router. I built responsive buyer and seller experiences, messaging, insights, marketplace discovery, and a photo-first listing wizard.
  • Backend: TypeScript and Express with secure cookie sessions, Zod validation, ownership checks, rate limiting, image processing, and transactional inventory updates.
  • Gemini: Generates editable drafts from item photos and powers Cora through structured tool calls. Server-side rules determine what actions Cora is authorized to take from marketing to customer-specific negotiation.
  • Solana: Phantom and @solana/web3.js enable buyer-approved test-SOL payments on devnet after a reservation is approved.
  • Tiger Data: Managed PostgreSQL stores accounts, listings, conversations, offers, reservations, notifications, and recommendation signals.
  • Vultr: Hosts the production application behind Caddy with HTTPS and a custom domain.
  • GoDaddy: Registered Domain (yourencore.work)

🧩Challenges I ran into

  • The fight for autonomy in the world of agentic AI. How much do we want our agents to dictate our work?
  • Devpost Blocker: Mid-way through the hackathon, DevPost had accidentally flagged and froze my account from making this Devpost that you’re reading right now. If it were not for the help I received from the organizers who were active and responsive throughout; had I not planned ahead for these unforeseen circumstances, Encore may not have lived to tell the tale.
  • Learning against the 36-Hour Clock: Tinkering with new tools like Solana, Tiger Data, and Vultr with little to no prior experience with Postgresql and Cloud Computing gave me an appreciation for the remarkable progress AI has made in making tech accessible for learning developers.

🏆Accomplishments that I'm proud of

  • Built and Deployed a Fully Functional E-Marketplace Solo: Took Encore from idea to a live, full-stack product.
  • Created a real AI Workflow: Cora answers, negotiates, escalates, and protects seller control.
  • Connected the Full Stack: Gemini, Tiger Data, Solana, and Vultr each serve a clear product purpose.
  • Handled Real marketplace Complexity: Private price limits, competing reservations, quantities, payments, recommendations, and notifications.
  • Designed for Trust: Unlike many AI agents, Cora admits uncertainty instead of inventing answers.
  • Made it Feel Cohesive: Built the product, brand, interface, animations, and responsive experience from scratch.

📚What I learned

  • I am capable: By solo-ing this project, I got to bring this project from idea to a fully functional product, overseeing all aspects of full-stack development and tinkering with tools I have never even heard of. More importantly, however, I learned to trust in my own abilities and embrace the challenges that help me grow.

  • The simplest solutions are the best solutions to the most complex of problems: In software and product development especially, a complex solution only complicates a problem even more by creating another problem for users -- how to navigate/use the solution. The most effective solutions are those that provide the highest utility in the most intuitive way. In creating a market place for ordinary individuals, not merchants, not businesses, this was paramount to my project.

  • The fight for autonomy in the world of agentic AI: How much do we want our agents to dictate our work? The context of the problem at hand should always lead the answer to this question. While Cora is there to automate tasks, it does not own the item like the seller does and cannot arbitrarily make decisions or invent details it does not know. Hence why sellers must verify the generated details before proceeding, why Cora escalates questions it cannot answer, and why sellers get the final say in approving a sale. AI does not replace human intention.

  • The importance of diversifying your tech stack: Simple. By using a variety of tools, if something unexpectedly breaks, it doesn't break your entire project.

🔮What's next for Encore

Next, I want to pilot Encore with 10–20 campus users and measure listing time, AI correction rates, resolved conversations, escalation accuracy, reservations, and completed exchanges.

Future improvements include stronger moderation, account recovery, expanded agent evaluations, mobile-first capture, better location discovery, and production-ready payment options.

The goal remains simple: make giving something a second life easier than leaving it forgotten.

Built With

Share this project:

Updates

Submission history