Home · AI Agents
Copy a prompt, point your agent at the LandlordIQ MCP tools, and automate the rental-analysis work you'd otherwise do by hand. Each playbook pairs a ready-to-paste prompt with the exact API call behind it. Estimated fields are labeled so your agent knows how far to trust each number.
Every playbook works two ways. If you've connected the MCP server to Claude, Cursor, or another client, paste the prompt and the agent will pick the right tools. If you're scripting against the HTTP API directly, copy the call instead. Both return the same envelope: data, confidence_score (0–100), data_freshness, data_source, and error.
Reminder on confidence: 72–92 is HUD-anchored or curated (high), 50–71 is an estimated blend (moderate), and below 50 is a national fallback or extrapolation (low). Tell your agent to present low-confidence numbers as ranges, not facts. See the confidence guide.
Hand the agent a candidate deal. It pulls the market context, benchmarks the underwritten cap rate against the metro band, and tells you whether the deal clears your hurdle — with sources and confidence attached.
For an existing tenant, the agent returns a recommended renewal rent, the gap to market, and a retention-risk read you can drop into a renewal letter. It cross-checks the recommendation against the raw market-rent estimate.
Run this on a schedule across the ZIPs you own. The agent checks rent-growth and vacancy shifts in each market and surfaces only the ones that crossed a threshold worth a human's attention — turning a portfolio review into an exception report.
Comparing where to deploy capital next? Give the agent a shortlist of ZIPs and it builds a side-by-side table of rent, growth, vacancy, and cap-rate band, then ranks them against your stated objective (yield, growth, or stability).
Before you set a rent increase, late fee, or notice period, the agent pulls the plain-English landlord-tenant summary for the city and flags anything you need to confirm with counsel. Output always carries the not-legal-advice disclaimer.
confidence_score < 50 as a range, not a point, and to recommend a human check.data_source and data_freshness next to every figure so you know the as-of date.get_landlord_market_summary is one call instead of four and keeps your rate-limit budget low.error or fallback: true, the agent should report it, not fabricate a number.