streeteasy-mcp
A remote MCP server that wraps the StreetEasy GraphQL API to enable LLM agents to search and parse NYC rental listings.
README
streeteasy-mcp
A remote MCP server that wraps the StreetEasy GraphQL API so an LLM agent can search and parse NYC rental listings.
It vendors the streeteasy-api
client (v0.4.0) and exposes it over either stdio (local) or Streamable
HTTP (remote) transport, so it can be connected to by Claude or any MCP client.
[!IMPORTANT] StreetEasy blocks datacenter/cloud IPs. Its API sits behind PerimeterX bot-detection that
403s cloud/datacenter IPs (AWS, GCP, Railway, etc.). The HTTP build deploys fine and the MCP layer works, but the upstreamsearch_rentals/get_rental_detailscalls fail from a datacenter unless you do one of:
- Run the stdio server locally from a normal home connection — see Run as a local MCP server, or
- Route upstream calls through a residential proxy by setting
STREETEASY_PROXY— this is what lets the hosted HTTP build (Railway, etc.) reach the API. See Proxy / bot-detection.
Tools
| Tool | Description |
|---|---|
search_rentals |
Search active NYC rentals by area, price, beds, baths, amenities, pets. Returns compact listings + totalCount, paginated. Each listing includes leadPhotoUrl / photoUrls and a listing url. |
get_rental_details |
Full detail for one listing id: description, amenities, pricing history, building info, nearby transit/schools, and resolved media — media.photoUrls, media.floorPlanUrls, media.videoLinks (YouTube/Vimeo), media.tour3dUrl. |
list_areas |
Look up StreetEasy area names ↔ numeric codes (optionally filtered by a search term). |
list_amenities |
List the valid amenity enum tokens. |
search_rentals accepts area names ("MANHATTAN", "Williamsburg",
"upper east side") or numeric codes, and validates amenity tokens against the
known set.
Media
Photos resolve to Zillow's CDN (photos.zillowstatic.com/fp/{key}-se_large_800_400.jpg),
videos to their provider watch URL (YouTube/Vimeo) plus a thumbnail, and 3D
tours to a direct tour3dUrl. All are public — no auth required.
Not included: contact info & inquiries
Listing agent contact details and "request a tour" inquiries are not exposed.
They live behind StreetEasy's contact flow, which is protected by PerimeterX
bot-detection (a "Press & Hold" human check). Automating it would mean evading
bot-detection, so it's intentionally left out — the right pattern is to surface
the listing url and let a human submit the tour request in their browser.
Endpoints
POST /mcp— the MCP Streamable HTTP endpoint (stateless). Requires a bearer token unlessMCP_DISABLE_AUTHis set — see Authentication.GET /andGET /health— health checks.- OAuth:
/.well-known/oauth-authorization-server,/.well-known/oauth-protected-resource,/register,/authorize,/token,/revoke.
Run as a local MCP server (recommended)
Runs over stdio from your machine's residential IP — the configuration that actually reaches StreetEasy.
npm install
npm run build
# register with Claude Code (uses the stdio entry point):
claude mcp add streeteasy -- node "$(pwd)/dist/stdio.js"
Then ask Claude to search rentals. To run the stdio server by hand:
npm run start:stdio
Run as an HTTP server
npm install
npm run build
npm start # listens on $PORT (default 3000), POST /mcp
Test it with the MCP SDK client (see test-client.mjs):
MCP_URL=http://localhost:3000/mcp node test-client.mjs
Configuration
| Env var | Purpose |
|---|---|
PORT |
Port to listen on. Railway sets this automatically. |
PUBLIC_BASE_URL |
Public origin the server is reachable at (the OAuth issuer), e.g. https://streeteasy-mcp-production.up.railway.app. Defaults to https://$RAILWAY_PUBLIC_DOMAIN on Railway, else http://localhost:$PORT. |
MCP_DISABLE_AUTH |
Set to 1/true to disable OAuth and leave /mcp open (handy for local testing with the bundled test-client.mjs). |
STREETEASY_PROXY |
Optional. HTTP/HTTPS proxy for all upstream StreetEasy calls, e.g. http://user:pass@host:port. Required for cloud/datacenter deploys — use a residential proxy. HTTPS_PROXY / ALL_PROXY are also honored. |
Authentication (OAuth 2.1 + Dynamic Client Registration)
The HTTP transport requires OAuth by default — MCP clients (Claude, etc.) run the standard authorization flow automatically, so you usually don't configure anything. The server is a self-contained OAuth 2.1 authorization server:
- Advertises metadata at
/.well-known/oauth-authorization-serverand/.well-known/oauth-protected-resource. - Supports Dynamic Client Registration (RFC 7591) at
/register, so clients self-register with no manualclient_id/client_secret. /authorize(PKCE S256 required) →/token(authorization-code + refresh), with/revokefor revocation.- Unauthenticated
POST /mcpreturns401with aWWW-Authenticateheader pointing at the protected-resource metadata, which kicks off discovery + DCR.
Because the tools expose only public listing data, there's no per-user login:
authorization is auto-approved and the issued bearer token simply gates /mcp.
Tokens are held in memory (single replica); a restart just makes clients
transparently re-register. Set MCP_DISABLE_AUTH=1 to turn the whole layer off.
Proxy / bot-detection
StreetEasy 403s datacenter IPs, so any cloud host (Railway included) must send
upstream requests through a residential proxy. Set STREETEASY_PROXY to a
proxy URL (credentials may be embedded, e.g.
http://user:pass@host:port) and all StreetEasy GraphQL traffic is tunnelled
through it. On startup the server logs the proxy in use with credentials
redacted (Outbound proxy: http://***:***@host:port).
Rotating residential proxies hand out a fresh exit IP per connection, and a
clean IP isn't guaranteed every time, so the client automatically retries a
403 bot-challenge (up to 3 times when a proxy is set) to land on a good IP.
A local stdio server on a residential connection doesn't need a proxy.
Deploy on Railway
This repo ships a Dockerfile. With the Railway CLI:
railway login
railway init --name streeteasy-mcp
# Cloud hosts are datacenter IPs — set a residential proxy so calls aren't 403'd:
railway variables --set "STREETEASY_PROXY=http://user:pass@host:port"
railway up
railway domain # generate a public URL
# Set the OAuth issuer to your public URL (or rely on RAILWAY_PUBLIC_DOMAIN):
railway variables --set "PUBLIC_BASE_URL=https://<your-app>.up.railway.app"
Connect from Claude Code
claude mcp add --transport http streeteasy https://<your-app>.up.railway.app/mcp
The client discovers the OAuth endpoints and registers itself automatically
(Dynamic Client Registration) — no client_id / token to configure.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.