MCP server

Connect Claude, Cursor, or ChatGPT to your Lucanto workspace

Lucanto ships a built-in Model Context Protocol server, so AI agents can read and act on your workspace data — list and create invoices, look up contacts, mark invoices paid, send them — using the same API key and scopes as the REST API.

Endpoint: POST https://app.lucanto.eu/api/mcp/v1
Transport: Streamable HTTP (stateless)
Auth: Authorization: Bearer lct_pat_... (a personal access token)

Claude Desktop

Add Lucanto to claude_desktop_config.json:

1 {
2   "mcpServers": {
3     "lucanto": {
4       "url": "https://app.lucanto.eu/api/mcp/v1",
5       "headers": {
6         "Authorization": "Bearer lct_pat_xxxxxxxxxxxx"
7       }
8     }
9   }
10 }

Restart Claude Desktop, then ask: “List my Lucanto workspaces” — the agent will call lucanto_list_workspaces and go from there.

Tools & scopes

The tool catalog is filtered by your token’s scopes — a read:invoices token never sees write tools. Each tool carries behavior hints so clients prompt appropriately (e.g. lucanto_delete_invoice is marked destructive, lucanto_send_invoice reaches outside Lucanto).

Scope tier	Example tools
read	`lucanto_list_workspaces`, `lucanto_list_invoices`, `lucanto_get_invoice`, `lucanto_list_contacts`, `lucanto_get_invoice_pdf_url`
write	`lucanto_create_invoice`, `lucanto_update_invoice`, `lucanto_create_contact`
manage	`lucanto_issue_invoice`, `lucanto_mark_invoice_paid`, `lucanto_send_invoice`, `lucanto_delete_invoice`

Start with a narrowly-scoped token. Give an agent read:* to explore, and only grant write/manage scopes once you trust the workflow.

1	{
2	"mcpServers": {
3	"lucanto": {
4	"url": "https://app.lucanto.eu/api/mcp/v1",
5	"headers": {
6	"Authorization": "Bearer lct_pat_xxxxxxxxxxxx"
7	}
8	}
9	}
10	}