RevNest

A Transparent and Agentic Revenue Management System for Independent Hosts

Inspiration

Enterprise grade pricing has traditionally been locked behind expensive analytic software and dedicated revenue management teams. For small business owners, it is incredibly hard to compete with corporations. We want to democratize the intricate balance between supply and demand for everyday Airbnb hosts and small hotel operators. This demographic is often the most busy and in need of a financial operator, and though the market has automated solutions- the expenses of those solutions are often out of reach for business owners operating small scale. We built RevNest to bridge the gap between an agent that is not only smart but entirely transparent.

What it does

RevNest acts as an autonomous revenue manager, the agent “Revy” is core to its design. It monitors changes in market conditions and recommends optimized rates semi-hourly. Revy doesn’t merely guess the price in a rapidly changing market, it actively fetches live data through APIs such as Ticketmaster Discovery API and Tavily Search API to fetch local weather, upcoming events, holiday schedules, competitor rates, and tourism search trends. Revy will then examine and determine reasons for a price change, and clarify to the user her exact train of thought. The host can then push the pending task with a manual approval or allow Revy to auto publish prices.

How we built it

We designed a decoupled architecture featuring a Next.js/React frontend and a robust Python-based agent backend.

The Brains: The core reasoning is powered by OpenClaw and NVIDIA's Nemotron models (nvidia/nemotron-3-super-120b-a12b).

The Operation Loop: When a pricing run triggers, the agent verifies property context (using headless browser scraping via MCP), and then dispatches parallel Python scripts to gather data from multiple APIs simultaneously to reduce latency.

The Data Layer: The agent synthesizes this data, calculates the RevPAR (Revenue Per Available Room) lift, and executes a secure write-back directly to a PostgreSQL database, which updates the Next.js dashboard instantly via Server-Sent Events (SSE).

Challenges we ran into

Trust & Safety in Financial AI: Letting an AI set prices is risky. We had to build a strict mathematical guardrail system that physically clamps suggested prices so they never violate host-defined minimums, maximums, or maximum week-over-week percentage changes.

Data Privacy: We had to ensure the agent didn't leak Personally Identifiable Information (PII) to external search tools. We solved this by implementing a secondary LLM script acting as a "Security Auditor" to intercept and block any PII (like guest names or exact revenue history) from leaving the secure environment.

Latency: Calling 5+ external APIs for market data sequentially was too slow. We had to implement a concurrent fan-out/fan-in Python orchestration script using ThreadPoolExecutor to run them in parallel.

Accomplishments that we're proud of

We are incredibly proud of our Guardrail System and the Live Reasoning Stream. By showing the host the exact math and market signals behind every decision, and explicitly proving that the AI is obeying their financial boundaries, we transformed a potentially scary "black-box" AI into a trusted financial co-pilot.

What we learned

We learned how to orchestrate complex, multi-stage LLM workflows (Directed Acyclic Graphs of prompts) where an agent is constrained to specific sub-skills (like gathering context, reviewing guardrails, and formatting outputs) rather than relying on one massive, error-prone prompt. We also learned how to leverage MCP (Model Context Protocol) to give our agent secure browser and data-fetching capabilities.

What's next for RevNest

We plan to expand RevNest's capabilities by adding direct API integrations with major Property Management Systems (PMS) like Guesty and Hostaway, allowing for direct price publishing without needing a database middleman. We also want to introduce deeper customization for host preferences, such as aggressive weekend pricing or long-stay discounts.