Connect AI agents (MCP)

The platform speaks the Model Context Protocol natively. Any MCP client — Claude Code, Claude Desktop, IDE agents, your own bots — can query availability, price stays and read bookings with live data, authenticated by the same API key and scopes as the REST API.

• Endpoint: https://api.myrentalcalendar.com/mcp
• Transport: Streamable HTTP, stateless (plain JSON responses, no SSE session)
• Auth: Authorization: Bearer hr_live_…

Claude Code

claude mcp add --transport http stayos https://api.myrentalcalendar.com/mcp \
  --header "Authorization: Bearer hr_live_XXXXXXXX"

Generic MCP client configuration

{
  "mcpServers": {
    "stayos": {
      "type": "http",
      "url": "https://api.myrentalcalendar.com/mcp",
      "headers": { "Authorization": "Bearer hr_live_XXXXXXXX" }
    }
  }
}

Tools

Only tools covered by the key’s scopes are advertised — a read:units-only key exposes nothing but list_units.

Tool	Scope	What it does
list_units	read:units	All units with slug, capacity, base price
get_availability	read:availability	Blocked nights of one unit in a range
quote_stay	read:rates	Full price quote incl. fees, taxes, promotions + availability
list_bookings	read:bookings	Booking rows (no guest personal data)

All tools are read-only; there is deliberately no tool that creates or modifies bookings. Argument mistakes come back as readable tool errors (for example the effective minimum stay), so agents can correct themselves.

Trying it with curl

MCP is JSON-RPC over HTTP — you can smoke-test the endpoint without any client:

curl -s https://api.myrentalcalendar.com/mcp \
  -H "Authorization: Bearer $STAYOS_KEY" -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"tools/list"}'