PropertyAgent

The true Agent Assisted Property Platform (Minus the Agent)

Comment

Inspiration

Everyday people in Singapore want to manage their own housing and rental transactions to save thousands in traditional agent commissions (1% to 2% for buyers/sellers; up to a month's rent for tenants). However, when they go DIY, they face an overwhelming information and compliance barrier. Current platforms act as disconnected silos, forcing users to manually synthesize fragmented market transaction indexes, regulatory timelines, and financial rules on their own. We built PropertyAgent to eliminate this manual friction, transforming a passive directory into an active, predictive transaction workspace that gives users complete autonomy.

What it does

PropertyAgent is a predictive C2C real estate platform that serves as a high-confidence digital co-pilot. Starting with an open-ended lifestyle conversational search bar, it guides buyers, sellers, and tenants through their entire transaction lifecycle. For buyers, it uses historical transaction data to calculate potential Cash Over Valuation (COV) cash traps upfront. Throughout the journey, the dashboard locks users into synchronized statutory milestone checklists, ensuring they never miss high-stakes regulatory deadlines.

How we built it

The platform is engineered using Next.js, serverlessly deployed on google cloud run. Users describe what they are looking for in natural language; the agent (Gemini 2.5 Flash via Google ADK), leverages the Elastic MCP server to build various Elasticsearch queries that enable our smart search functionality. The agent also has access to Google calendar tools that enables it to make viewing appointments for users.

Challenges we ran into

  • Agent builder: the UI console based agent builder did not support authentication for MCP servers, necessitating the use of the ADK instead, resulting in more code to maintain
  • Familiarity with the Google cloud stack: As frequent users of other cloud systems, it was interesting to dive deep into the google cloud stack. As it was designed to be easy to pick up and use, and we got up to speed quickly

Accomplishments that we're proud of

  • Context-Aware Lifestyle Search: Moving away from restrictive pre-set filters to an open-ended semantic bar that successfully maps fluid human desires (e.g., "near a park, short commute to Raffles Place") into rigid geographic and structural search boundaries.
  • Intelligent Proxy Reasoning: Leveraging Gemini's native reasoning capabilities to gracefully infer and recommend smart proxy locations when Elasticsearch returns empty sets for low-volume blocks. The model intelligently identifies and evaluates structural and situational similarities across adjacent neighborhoods entirely through semantic text logic.

What we learned

We learned that when it comes to high-stakes real estate transactions, transparency is more important than absolute prediction. Users do not expect a digital tool to be a literal crystal ball that knows the interior renovation quality of a flat; they expect it to be a highly calibrated historical anchor. We realized that framing data limitations honestly using clear UI clue text and confidence scores builds significantly more product trust than over-promising automated accuracy.

What's next for PropertyAgent

  • Upstream User Verification Integration: Building out the core security layer to verify users via Singpass. This includes developing the automated financial pre-qualification filter that intercepts unverified accounts and requires them to link their HFE parameters before they can unlock a direct seller's booking calendar.
  • Seamless, Data-Backed Bidding System: Developing a structured, real-time bidding experience designed to lower emotional friction during the offer stage. The interface will replace anxious, gut-feel bidding with transparent price intelligence, matching active offers against localized percentile distribution graphs and immediate cash outlay calculations.
  • Synchronized Step-by-Step Milestone Guides: Building an automated, interactive timeline tracker that maps out the complete statutory transaction journey. The platform will look users into synchronized checklists based on official HDB and CEA requirements, tracking milestones from option intent down to final legal completion so users never miss a critical administrative or payment deadline.

Built With

Share this project:

Updates