• 

  Logo

Rent-Swarm

The Autonomous Tenant Defense System

Inspiration

Rent-Swarm was inspired directly by our own housing struggles as students in competitive city markets. Farhan, a University of Maryland student, is expected to move off campus in his second year—like most UMD students—and is already facing a highly competitive housing market. Maanya studies at UT Austin, where demand has surged faster than housing development, making affordability a constant concern. Oluwaferamni attends Howard University in Washington, DC, one of the most expensive and in-demand rental markets in the country. Beyond our team, we have close friends in Boston-area universities and at UIUC who are currently struggling to secure housing in oversaturated markets.

Across these cities, we noticed the same pattern: landlords operate with automation, data, and leverage, while tenants are left to manually search fragmented listings, read dense leases, and negotiate blindly. We wanted to build a system that gives renters—especially students—the same algorithmic power.

What it does

Rent-Swarm is an agentic AI platform that acts on behalf of tenants. It autonomously scouts housing listings, flags likely scams, analyzes lease agreements for dangerous or illegal clauses, calculates the true cost of living, and drafts professional rent negotiation emails using comparable listings as leverage.

How we built it

Rent-Swarm is designed as a multi-agent system, where each agent performs a specific form of labor. A scouting agent navigates housing platforms using Browserbase to identify listings and detect scams. A legal analysis agent uses large language models with retrieval-augmented generation (RAG) using AWS Bedrock to analyze lease documents against local tenant protection laws. A financial agent computes the Real Effective Monthly Cost using:

$$ \text{REMC} = \text{Base Rent} + \text{Estimated Utilities} + (\text{Commute Cost} \times 20) + \text{Hidden Fees} $$

A negotiation agent then synthesizes this information to draft professional, data-backed negotiation messages. The backend is orchestrated using LangGraph, while the frontend is built with React and Next.js for a fast, transparent user experience.

Challenges we ran into

We initially planned to display real housing listing images directly from platforms, but technical limitations made reliably obtaining and rendering them within the hackathon timeframe difficult. To preserve usability and visual quality, we used high-quality placeholder home images to keep the interface dynamic and aesthetically polished.

We also included to show a live video stream of Browserbase agents navigating the web in real time.

Accomplishments that we're proud of

We built a fully functional and polished frontend within the first few hours of the hackathon, which allowed us to quickly validate the user experience and focus the remainder of our time on refining the idea and backend logic. We’re especially proud of how clearly Rent-Swarm demonstrates agentic labor—searching, reading, reasoning, and writing autonomously on behalf of the user.

What we learned

We learned how powerful agent-based systems can be when tasks are clearly separated and carefully orchestrated. Designing agents that perform real work requires thoughtful coordination, robust logic, and trust-building through UX.

We also learned the value of prioritizing frontend design early, as a strong interface accelerated iteration and clarified our product vision.

What's next for Rent-Swarm

Next, we plan to expand Rent-Swarm with secure lease signing, background checks for landlords, tenants, and roommates, and neighborhood safety scores based on public data. Our long-term goal is to make renting in metropolitan areas transparent, fair, and tenant-first.

Built With

• amazon-web-services
• browserbase
• gemini
• langchain
• llamaindex
• rag
• react
• tailwindcss
• typescript
• v0