Connect Claro to an MCP client

Claro AI is a platform product. Its MCP server is at https://platform.casafari.com/mcp/claro: Streamable HTTP with JSON responses, stateless, tools only. It signs in with a platform API key or the platform's OAuth, so a client that handles OAuth (claude.ai, ChatGPT) needs only the URL, and any other client sends the key in an Authorization header. The platform meters each tool call in platform units and forwards it to Claro unchanged; Claro's Claro-Version header and error bodies pass through. Create a platform API key in your account on platform.casafari.com. In every block below, replace $CASAFARI_API_KEY with your key, or keep it in an environment variable or the client's secret store; never commit it. What Claro is: Claro AI. The Casafari MCP gateway has its own page: Connect an MCP client.

#Claude Code

claude mcp add --transport http claro https://platform.casafari.com/mcp/claro --header "Authorization: Bearer $CASAFARI_API_KEY"

Then run /mcp inside Claude Code: claro is listed with its tools. To sign in with OAuth instead of a key, leave out --header and authenticate from /mcp.

#Claude desktop and claude.ai

A custom connector in claude.ai (Settings → Connectors → Add custom connector) takes a URL and signs in with OAuth. Through the platform's OAuth it can reach Claro: paste https://platform.casafari.com/mcp/claro and sign in with your platform account. Claude desktop uses the same connectors.

#ChatGPT

In ChatGPT's settings turn on Developer mode (Apps & Connectors, advanced settings), create a connector, paste https://platform.casafari.com/mcp/claro as the MCP server URL and sign in with the platform's OAuth. From your own code, the OpenAI Responses API sends the key for you:

{
  "model": "gpt-5",
  "input": "burger restaurant, corner unit 120-250 m², smoke outlet, busy street, Madrid, rent under 8,000 a month",
  "tools": [
    {
      "type": "mcp",
      "server_label": "claro",
      "server_url": "https://platform.casafari.com/mcp/claro",
      "authorization": "$CASAFARI_API_KEY",
      "require_approval": "never"
    }
  ]
}

#Cursor

Add to ~/.cursor/mcp.json (or .cursor/mcp.json in a project):

{
  "mcpServers": {
    "claro": {
      "url": "https://platform.casafari.com/mcp/claro",
      "headers": { "Authorization": "Bearer ${env:CASAFARI_API_KEY}" }
    }
  }
}

#VS Code

Add to .vscode/mcp.json. VS Code asks for the key once and keeps it in its secret store:

{
  "inputs": [{ "type": "promptString", "id": "casafari-key", "description": "Casafari platform API key", "password": true }],
  "servers": {
    "claro": {
      "type": "http",
      "url": "https://platform.casafari.com/mcp/claro",
      "headers": { "Authorization": "Bearer ${input:casafari-key}" }
    }
  }
}

#Any other client

Point the client at https://platform.casafari.com/mcp/claro and have it send Authorization: Bearer $CASAFARI_API_KEY with every request. To check the connection by hand:

curl -X POST "https://platform.casafari.com/mcp/claro" \
  -H "Authorization: Bearer $CASAFARI_API_KEY" \
  -H "Content-Type: application/json" \
  -H "Accept: application/json, text/event-stream" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'

The answer lists the 9 tools: search_premises, refine_search, get_premises, compare_premises, market_snapshot, explain_location, brief_from_file, search_plots, get_plot. Without a valid platform key or token the platform answers 401.

#Check it works

Ask your assistant: "burger restaurant, corner unit 120-250 m², smoke outlet, busy street, Madrid, rent under 8,000 a month" It should call search_premises and answer with counts, the criteria it understood and the premises that match, each with the sentence it rests on and the agency's own page.