Tab 05 · Connectors

The databases, inside Claude.

The supply, rent and demand databases are exposed to Claude as read-only MCP tools. Connect once, then ask in plain English and the server runs the query. A free public connector covers site screening for anyone; the full server is what I use for underwriting.

§ 01

Rental Market Data

Multifamily & BFR · free · public · no account

A read-only connector that screens any U.S. site for build-to-rent and multifamily: what is permitted or entitled nearby, what comparable rentals are asking after concessions, who lives within a few miles, and how the ZIP’s home values compare with its metro.

Add to Claude Docs, example prompts and limits Privacy policy Terms Support

Codex: the button copies the server URL and opens OpenAI’s Codex setup guide in a new tab. In Codex open Settings → MCP servers → Add server, choose Streamable HTTP, paste the URL and save, then select Authenticate.

Server URL: https://mcp.hermanchan.ai/mcp

Public or private

	Public connector	Private server
Who	Anyone with Claude	By invitation
Sign-in	One consent click, no account	OAuth 2.1 or a per-person bearer token
Tools	Nine fixed tools, one address at a time	Seventeen tools, including search, builder activity and source documents
Data	Public records, operator rents, Census, Zillow ZHVI	The full supply, rent and demand databases

§ 02

Ask the databases

Live · try it here, nothing to install

The private server below, with Claude in front of it. Every answer shows which methods it called and which records it cited; a question the databases cannot answer says so rather than guessing.

Ask the supply, rent and demand databases — one server holds all three.

This box runs the same Model Context Protocol server I query from claude.ai: nine metros of supply, the weekly rent scrape, and Census demand rings. Five single-database questions, three that cross them, and one it should decline. Click a question to ask it.

Guided tour · each chip names the MCP methods it calls

20 questions left today · the server also limits questions per hour · answers are grounded in live data from the three databases; verify before acting.

§ 03

The private MCP server

Read-only · OAuth 2.1 · all nine metros

The full databases, exposed to Claude as a small set of read-only tools. I connect it to claude.ai once and then ask in plain English — "what has been filed near this site in the last ninety days", "what is this builder doing in Atlanta" — and the server runs the query and returns a clean answer. The chat box above uses the same tools.

Endpoint: https://permit-pipeline-mcp-production.up.railway.app/mcp
Auth: OAuth 2.1 (claude.ai custom connector) or a bearer token issued per person. Access is by invitation; the tools are read-only against the database.
Connect: In claude.ai: Settings → Connectors → Add custom connector → paste the endpoint → authorise. Claude Code: claude mcp add --transport http supply https://permit-pipeline-mcp-production.up.railway.app/mcp.

Tool	What it answers
`find_subdivisions`	Search the pipeline with structured filters: metro, product, stage, size, date.
`recent_filings`	What was filed across the pipeline in the last N days.
`lookup_subdivision`	Look up one project by name or by any of its record IDs.
`permits_near`	Filings within a radius of a point in the last N days.
`off_market_signals`	Projects building but not yet selling — the pre-listing pattern.
`builder_activity`	One builder's activity across every source, with alias resolution.
`canonical_project`	The full cross-source timeline for one project.
`get_pdf_url`	A signed link to the stored source document for a filing.
`pipeline_health`	Ingestion freshness per metro and per source.
`mf_projects`	Late-stage multifamily and attached projects from the permit layer.
`entitlement_leads`	Pre-permit leads: rezonings, PUDs, site plans, plats, district formations.
`debt_maturities`	Financed multifamily with loan maturities and the permitted units within three miles. Historical: its loan data stopped updating on September 7, 2026.
`rent_comps_near`	Rent database: floor-plan-level asking and net-effective rents within a radius of a point.
`rent_market_summary`	Rent database: market medians, concession share and depth, availability — overall and by bed count.
`rent_community_lookup`	Rent database: a community by name, with its coordinates and current floor plans.
`rent_coverage`	Rent database: which markets are covered and how fresh each is.
`demand_rings`	Demand: population, households, income and jobs in 1/3/5/10-mile rings around an address, from Census and LODES.