Casafari MCP

Property data intelligence, built for AI agents. Casafari is the AI agent-native real estate data intelligence platform. Deduplicated, cleaned property graph. Europe’s most complete property index.

Connect Claude, ChatGPT, Cursor, VS Code, Gemini CLI or your own agent to five MCP servers on one account: comparables and price estimates, area insights, location lookup, agency analysis in Germany, and Claro AI, which finds commercial premises from a brief in plain words. Sign in with OAuth in the browser; there is no key to copy.

/plugin marketplace add casafari/casafari-mcp
/plugin install casafari@casafari

That is the whole setup in Claude Code. Every other client is below.

Now in Claude’s Connectors Directory

Claro AI is listed in Claude’s Connectors Directory: open it in Claude, click Add, sign in with your Casafari platform account, and ask.

Now in the Cursor Directory

Casafari is listed in the Cursor Directory: open it, click Add to Cursor next to a server, and sign in with your Casafari platform account.

Claro AI in Claude: real sessions

Three moments from real Claude sessions with the Claro AI connector, recorded on 16 October 2026.

Claude sees the tools. Ask what Claro can do and Claude lists its ten tools, from a brief in plain words to Location Intelligence for any point in Spain.

A brief becomes a ranked shortlist. “Specialty coffee shop, retail for rent in Valencia, 40-120 m², street level with shop window, cafetería licence, busy street or near offices and universities, under €3,500 a month.” Claro finds 546 matching premises, and Claude walks through the best six with rent, size, licence, footfall, residents and the cafés already nearby.

A committee-ready report, in one conversation. For a mid-size fitness club across the Madrid and Barcelona metros, Claude screens 483 matches down to seven candidates, scores them, and writes a dossier per site: the case for it, the figures behind it and what due diligence must confirm.

Example reports

All six are also on the developer hub, with a short brief for each: developers.casafari.com/docs/claro/examples.

Each report was written by Claude from Claro AI data in a single chat. Open either one in your browser; each is one self-contained page.

Edited for publication only to remove portal names and a client’s name. Figures are Claro AI’s: asking prices; footfall is an index for comparing places, not a head count.

One brief, one honest answer

A real call to Claro AI, run on 16 October 2026:

“Specialty coffee shop in Barcelona, street level, 60-120 m², smoke outlet, rent under 4,000 euros a month”

Claro reads the brief into criteria, checks every commercial premises for rent in Barcelona against each one, and counts exactly:

The top result, as the agent receives it:

Shell unit in La Marina de Port, Barcelona · 98 m² · €1,650 a month ✅ price · ✅ size · ✅ street level · ✅ smoke outlet · ✅ suited to a café Facts read from its own text: shop window, 6.22 m frontage, 3.74 m ceiling. Around it: footfall index 99,070 within a 5-minute walk (weekdays 105,429, weekends 82,931), 24,511 residents within a 10-minute walk, 3 cafés nearby. Every figure carries its unit, source and vintage. The link goes to the agency’s own page.

And the same answer as data, ready to chart:

xychart-beta
    title "What each criterion leaves out (1,025 matches)"
    x-axis ["Suited to a café", "Smoke outlet", "Street level", "Price", "Size"]
    y-axis "Premises that state it" 0 --> 1100
    bar [262, 342, 387, 927, 1001]

xychart-beta
    title "Asking rents, retail 60-120 m², Barcelona (963 with a stated rent)"
    x-axis ["<1k", "1-2k", "2-3k", "3-4k", "4-5k", "5-6k", "6-7k", "7-8k", "8-9k", "9k+"]
    y-axis "Premises" 0 --> 520
    bar [235, 496, 162, 30, 14, 4, 6, 2, 0, 14]

Median asking rent €1,400 a month. 927 of 963 fit the €4,000 budget. Asking prices, each unit counted once.

How it fits together

flowchart LR
    A["Your agentClaude · ChatGPT · CursorVS Code · Gemini CLI · custom"] -- "OAuth 2.1or cf_sk_ key" --> P(("platform.casafari.com"))
    P --> L["Location Lookup/mcp/locations"]
    P --> C["Comparables/mcp/comparables"]
    P --> I["Area Insights/mcp/area-insights"]
    P --> G["AgentGraph AI/mcp/agentgraph"]
    P --> K["Claro AI/mcp/claro"]
    L & C & I & G & K --> D[("Casafari property graphone record per property,with its full history")]

One account, one key and one balance of units work on every server. Location Lookup is free and gives the location ids the other products take.

What you can ask

The five MCP servers

Each server has a server card at /.well-known/mcp/server-card.json. Tools are named _, and every server also has account_status.

MCP clients sign in with OAuth 2.1 in the browser: the server answers 401 with a WWW-Authenticate header that points to its protected-resource metadata, and a client discovers the rest from there. No key to copy. Unattended agents use a platform API key instead (cf_sk_..., as Authorization: Bearer or X-API-Key).

Connect

Connect Location Lookup first; it is free and gives the ids the other products take. Docs for every client: https://developers.casafari.com/docs/connect.md

Claude (claude.ai and desktop)

Settings, Connectors, add a custom connector, paste a server URL from the table (for example https://platform.casafari.com/mcp/comparables) and sign in. Repeat for each server you want.

Claude Code

Install the plugin from this repository (all five servers and five skills):

claude plugin marketplace add casafari/casafari-mcp
claude plugin install casafari@casafari

Or add servers one by one, then run /mcp to sign in:

claude mcp add --transport http casafari-comparables https://platform.casafari.com/mcp/comparables
claude mcp add --transport http casafari-area-insights https://platform.casafari.com/mcp/area-insights
claude mcp add --transport http casafari-locations https://platform.casafari.com/mcp/locations
claude mcp add --transport http casafari-agentgraph https://platform.casafari.com/mcp/agentgraph
claude mcp add --transport http casafari-claro https://platform.casafari.com/mcp/claro

Docs: https://code.claude.com/docs/en/plugins and https://code.claude.com/docs/en/mcp

Cursor

From the Cursor Directory, or one-click install links below.

One-click install, one link per server (format from https://cursor.com/docs/mcp/install-links):

• casafari-comparables
• casafari-area-insights
• casafari-locations
• casafari-agentgraph
• casafari-claro

Or put this in ~/.cursor/mcp.json (or .cursor/mcp.json in a project), schema from https://cursor.com/docs/mcp:

{
  "mcpServers": {
    "casafari-comparables": {
      "url": "https://platform.casafari.com/mcp/comparables"
    },
    "casafari-area-insights": {
      "url": "https://platform.casafari.com/mcp/area-insights"
    },
    "casafari-locations": {
      "url": "https://platform.casafari.com/mcp/locations"
    },
    "casafari-agentgraph": {
      "url": "https://platform.casafari.com/mcp/agentgraph"
    },
    "casafari-claro": {
      "url": "https://platform.casafari.com/mcp/claro"
    }
  }
}

This repository is also a Cursor plugin (.cursor-plugin/plugin.json, mcp.json, plugin/skills/), see https://cursor.com/docs/plugins.

VS Code

.vscode/mcp.json (schema: https://code.visualstudio.com/docs/copilot/customization/mcp-servers):

{
  "servers": {
    "casafari-comparables": {
      "type": "http",
      "url": "https://platform.casafari.com/mcp/comparables"
    },
    "casafari-area-insights": {
      "type": "http",
      "url": "https://platform.casafari.com/mcp/area-insights"
    },
    "casafari-locations": {
      "type": "http",
      "url": "https://platform.casafari.com/mcp/locations"
    },
    "casafari-agentgraph": {
      "type": "http",
      "url": "https://platform.casafari.com/mcp/agentgraph"
    },
    "casafari-claro": {
      "type": "http",
      "url": "https://platform.casafari.com/mcp/claro"
    }
  }
}

ChatGPT

Turn on Developer mode in ChatGPT’s settings (Apps and Connectors, advanced settings), create a connector, paste a server URL from the table as the MCP server URL, choose OAuth and sign in. Steps as published at https://developers.casafari.com/docs/connect.md.

Gemini CLI

Install the extension (all five servers and GEMINI.md):

gemini extensions install https://github.com/casafari/casafari-mcp

Then run /mcp auth casafari-comparables (and the others) inside Gemini CLI. Or add servers one by one; the entry lands in .gemini/settings.json as httpUrl:

gemini mcp add --transport http casafari-comparables https://platform.casafari.com/mcp/comparables
gemini mcp add --transport http casafari-area-insights https://platform.casafari.com/mcp/area-insights
gemini mcp add --transport http casafari-locations https://platform.casafari.com/mcp/locations
gemini mcp add --transport http casafari-agentgraph https://platform.casafari.com/mcp/agentgraph
gemini mcp add --transport http casafari-claro https://platform.casafari.com/mcp/claro

Docs: https://github.com/google-gemini/gemini-cli/tree/main/docs/extensions

Any other MCP client

Any client that supports remote MCP servers over Streamable HTTP with OAuth works: give it a server URL from the table.

REST quick start

Base URL https://platform.casafari.com/v1, OpenAPI at https://platform.casafari.com/v1/openapi.json (one per product at /v1//openapi.json, Claro at https://platform.casafari.com/v1/claro/openapi.json). Send the key as Authorization: Bearer or X-API-Key, never in a query string.

export CASAFARI_API_KEY=cf_sk_...   # from https://platform.casafari.com/console/

# Location Lookup (free)
curl -H "Authorization: Bearer $CASAFARI_API_KEY" \
  "https://platform.casafari.com/v1/locations/search?query=Madrid&country=ES"

# Comparables
curl -X POST https://platform.casafari.com/v1/comparables/search \
  -H "Authorization: Bearer $CASAFARI_API_KEY" -H "Content-Type: application/json" \
  -d '{"operation":"sale","comparables_types":["apartment"],"comparables_count":5,
       "location_boundary":{"circle":{"distance":1,"target_point":{"coordinates":{"latitude":40.4168,"longitude":-3.7038}}}}}'

Full examples in examples/: curl, Python and TypeScript, each a Location Lookup, a Comparables call and a Claro search.

The existing Casafari API at api.casafari.com is for contract clients only; its tokens do not work on the platform.

What is in this repository

Skills: casafari-comparables, casafari-area-insights, casafari-locations, casafari-agentgraph, casafari-claro.

Links

• Docs: https://developers.casafari.com (machine-readable: https://developers.casafari.com/llms.txt)
• Example reports: https://developers.casafari.com/docs/claro/examples
• Platform and console: https://platform.casafari.com
• Terms: https://developers.casafari.com/legal/platform-terms
• Contact: [email protected]
• Security: see SECURITY.md

Licensed under the Apache License 2.0.