Negotiagent

• 

• 

• 

• 

Inspiration

High-volume sellers on Facebook Marketplace spend hours daily negotiating with buyers. They miss opportunities from slow responses and lose money accepting lowball offers out of exhaustion. We realized AI could handle repetitive negotiations while maintaining a human touch, letting sellers focus on what matters 💪

What it does

Negotiagent is a Chrome extension that negotiates automatically on behalf of sellers. Sellers set: minimum price, negotiation strategy, discount limits AI handles: analyzing buyer messages, generating natural responses, counteroffering strategically Safety features: hard price floors, manual approval for edge cases, escalation for unusual requests Result: Close more deals at better prices while saving hours every day.

How we built it

Tech Stack:

Gemini AI for natural language processing

JavaScript/HTML/CSS for UI

Content scripts monitor Facebook Marketplace in real-time, background workers handle secure API calls to Gemini, and an intuitive popup interface lets sellers configure their negotiation strategy in seconds.

Architecture:

Content scripts monitor Facebook Marketplace messages in real-time Background workers handle secure Gemini API calls Popup interface for seller configuration Local storage for preferences and conversation state

Challenges we ran into

Facebook's dynamic interface - Complex DOM manipulation made message interception tricky Prompt engineering - Balancing firm pricing with friendly, human-like tone Edge cases - Handling images, off-topic questions, aggressive buyers Rate limiting - Optimizing API calls and implementing smart caching Solution: Built a confidence scoring system that only auto-responds when certain, otherwise escalates to seller.

Accomplishments that we're proud of

✔︎ Fully functional Chrome extension working on live Facebook Marketplace

✔︎ Natural conversations that sound human, not robotic

✔︎ 100% safety record - never breached price floors in testing

✔︎ 60-second setup - clean, intuitive UI

✔︎ Better outcomes - closes deals at higher prices, 10x faster response time

What we learned

Browser extensions - Deep dive into content scripts, security models, and permissions

Prompt engineering - Structured context and clear constraints are crucial for AI consistency

Safety first - Guardrails and boundaries prevent costly mistakes

When to say NO - Knowing when not to auto-respond is as important as responding

User testing reveals everything - Real seller feedback uncovered UX issues we never anticipated

What's next for Negotiagent

Short-term: User testing with 10-20 real power sellers Analytics dashboard (win rate, average discount, time saved) Buyer behavior profiling

Medium-term: Multi-platform support (OfferUp, Craigslist, eBay) A/B testing negotiation strategies Automated listing creation

Long-term vision: Complete seller automation suite with inventory management, predictive pricing, and market trend analysis.

---

Built With

• claude
• geminiapi