How do I connect Claude or ChatGPT to wlthy?

wlthy exposes a Model Context Protocol (MCP) server over Streamable HTTP at https://wlthy.io/mcp. Create a read-only token in Settings → API & MCP (Black plan), then point your agent at the endpoint with that token — via the mcp-remote shim for Claude Desktop, or a url + headers entry for Cursor. The agent can then answer questions from your live ledger.

Can the AI agent move money or change my wlthy account?

No. The token is read-only by default: the agent can query your portfolio and stage a document for your review, but it cannot delete data, change settings or move money. No money-movement surface exists anywhere in wlthy. Every call is logged and any token can be revoked in one click.

wlthy

Model Context Protocol

Connect your AI assistant to your wealth ledger.

Ask about your global assets, partner splits, or live net worth right inside Claude or ChatGPT. Read-only by default, cryptographically secured, and scoped to a token you control.

Built on wlthy's encrypted ledgerEvery answer traceable to your ledgerRead-only, never writes

MCP Endpoint (Streamable HTTP)

https://wlthy.io/mcp

Get in Black

01 · What you can ask

Real questions for your AI.

The interface is natural language. You don't need to know table names or JSON structures. Ask the way you'd ask your financial advisor.

NET WORTH

“What's my total net worth in EUR, and how much is liquid cash?”

ALLOCATION

“How am I allocated by country and asset class?”

PARTNER SPLITS

“List all real estate assets where my spouse holds more than 30% equity.”

AI PARSING

“Parse this attached PDF broker statement and stage the positions for my review.”

02 · The toolset

Seven capabilities in one server.

The server ties directly into your encrypted ledger. You never have to work out which endpoint to call; the model routes it automatically.

get_net_worth

Returns total net worth with 1-day and 30-day deltas, plus asset and debt totals and a count of anything still unvalued.

list_portfolios

Lists every portfolio in the account — id, name, base currency — so the agent can scope the rest of its questions.

list_assets

Fetch detailed positions (active or sold) with pagination and live FX rates. The sold lens returns exit price, fees and taxes.

get_asset

Full detail for a single holding plus its 30-day value history — for a deep dive on one position.

get_allocation

Granular breakdown by asset class, geography, or currency — the concentration view, on demand.

get_recap

Period-end balances (not naive sums) across net worth, class, sector or crypto — daily, monthly or quarterly.

parse_document

Hands the agent a statement (PDF / CSV / image) to stage extraction for your manual approval — nothing is saved until you confirm.

03 · Integration guide

Setup takes under a minute.

Create a read-only token in Settings → API & MCP (Black plan), then drop it into your client config below. Replace wlthy_mcp_YOUR_TOKEN with the token you minted.

Claude Desktop

Open Settings › Developer › MCP Servers, add the config below, then restart Claude. The mcp-remote shim bridges Claude's local transport to the HTTP endpoint.

claude_desktop_config.json

{
  "mcpServers": {
    "wlthy": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://wlthy.io/mcp",
        "--header",
        "Authorization: Bearer wlthy_mcp_YOUR_TOKEN"
      ]
    }
  }
}

Cursor IDE

Add wlthy to your global ~/.cursor/mcp.json. Cursor speaks Streamable HTTP natively — no shim needed. See the Cursor MCP docs.

~/.cursor/mcp.json

{
  "mcpServers": {
    "wlthy": {
      "url": "https://wlthy.io/mcp",
      "headers": {
        "Authorization": "Bearer wlthy_mcp_YOUR_TOKEN"
      }
    }
  }
}