Connect MyFinanceMCP to Your AI Assistant

Follow the steps below to connect this MCP server to Claude Desktop, ChatGPT, Cursor, or any MCP-compatible client. Once connected, you can track your Angel One portfolio and even place trades through your AI assistant.

Don't want to set up MCP? You can get AI-powered portfolio insights directly from the web dashboard — just log in and add your OpenRouter API key. No MCP configuration needed.

Before You Start

Your Angel One API Key, Client ID, PIN, and TOTP Secret — learn how to generate them
Node.js installed (required for Claude Desktop — download here). Not needed for ChatGPT or Cursor.

Claude Desktop

Find your config file

Open the Claude Desktop config file at:

macOS

~/Library/Application Support/Claude/claude_desktop_config.json

Windows

%APPDATA%\Claude\claude_desktop_config.json

Add the MCP server

Add the following to the mcpServers object in the config file:

claude_desktop_config.json

{
  "mcpServers": {
    "angelone-portfolio": {
      "command": "npx",
      "args": ["-y", "mcp-remote", "https://myfinancesmcp.onrender.com/mcp/sse"]
    }
  }
}

Note: Claude Desktop doesn't support SSE natively, so we use mcp-remote as a bridge. It is installed automatically via npx.

Restart Claude Desktop

Quit Claude Desktop completely and reopen it. You should see a hammer icon indicating MCP tools are available.

Login

In your first message, ask Claude to log in with your credentials:

Example prompt

Log in to my Angel One account with these credentials:
- API Key: YOUR_API_KEY
- Client ID: YOUR_CLIENT_ID
- Password: YOUR_PIN
- TOTP Secret: YOUR_TOTP_SECRET

Claude will call the login tool, and then you can ask things like "Show my portfolio summary" or "What are my holdings?"

ChatGPT

Open Settings

On chatgpt.com, click your profile icon and go to Settings → Connected apps.

Add an MCP server

Click "Add" and fill in the following:

Name: AngelOne Portfolio

Description: Angel One portfolio tracker via MCP

MCP Server URL:

MCP Server URL

https://myfinancesmcp.onrender.com/mcp/sse

Authentication: select None

Check the "I understand and want to continue" checkbox

Click Create

Note: MCP support is available on ChatGPT web (chatgpt.com). No desktop app or Node.js required.

Login & use

Start a new chat and ask ChatGPT to log in with your Angel One credentials, then ask about your portfolio, holdings, or place trades.

Cursor IDE

Add the MCP config

Create or edit .cursor/mcp.json in your project root:

.cursor/mcp.json

{
  "mcpServers": {
    "angelone-portfolio": {
      "url": "https://myfinancesmcp.onrender.com/mcp/sse"
    }
  }
}

Note: Cursor supports the url field directly — no bridge needed.

Login & use

Open Cursor's AI chat and ask it to log in with your Angel One credentials, then use any portfolio or trading tool.

Other MCP Clients

Any MCP-compatible client that supports SSE transport can connect using:

https://myfinancesmcp.onrender.com/mcp/sse

If your client only supports stdio, use mcp-remote as a bridge:

npx -y mcp-remote https://myfinancesmcp.onrender.com/mcp/sse

Run Locally (Self-Hosted)

Don't want to use our hosted server? Clone the repo and run everything on your machine — no data leaves localhost.

Clone & install

Terminal

git clone https://github.com/raghav18482/MyFinancesMCP.git
cd MyFinancesMCP
pip install -r requirements.txt
python main.py

The server starts on http://localhost:8000. Open it in your browser for the web dashboard, or connect your AI assistant using the local MCP endpoint.

Claude Desktop — direct Python config (no Node.js needed)

When running locally, you can skip mcp-remote entirely and point Claude Desktop straight at the Python server:

claude_desktop_config.json

{
  "mcpServers": {
    "angelone-portfolio": {
      "command": "python3",
      "args": ["/full/path/to/MyFinancesMCP/mcp_server.py"],
      "env": {
        "ANGELONE_API_KEY": "YOUR_API_KEY",
        "ANGELONE_CLIENT_ID": "YOUR_CLIENT_ID",
        "ANGELONE_PASSWORD": "YOUR_PIN",
        "ANGELONE_TOTP_SECRET": "YOUR_TOTP_SECRET"
      }
    }
  }
}

Replace the highlighted paths and credential values with your own. On macOS/Linux you can find the full Python path by running which python3 in your terminal. This config runs the MCP server as a subprocess of Claude Desktop using stdio transport — no extra dependencies needed.

Cursor & ChatGPT — local SSE

With python main.py running, use the local URL in the configs shown in Steps 1–3 above:

http://localhost:8000/mcp/sse

Available Tools

Once connected and logged in, your AI assistant can use these tools:

Tool	Description
`login`	Authenticate with Angel One (must call first)
`logout`	Discard session credentials from memory
`portfolio_summary`	Overall portfolio: invested, current value, P&L
`get_holdings`	All stock holdings with P&L per stock
`get_positions`	Open intraday / delivery positions
`get_funds`	Available cash, margin, and net value
`get_order_book`	Today's orders and their status
`get_trade_book`	Executed trades for today
`search_scrip`	Search stocks by name or symbol
`get_ltp`	Last traded price for a stock
`get_candle_data`	Historical OHLC candle data
`place_order`	Place a buy/sell order (real money!)
`modify_order`	Modify an existing open order
`cancel_order`	Cancel an open order

Security

Credentials are never stored on disk — they live only in server memory
Sessions auto-expire after 8 hours
All traffic goes over HTTPS when deployed (Railway/Render provide TLS)
Your OpenRouter API key (for dashboard AI insights) is encrypted client-side and never stored on the server
You can call logout at any time to immediately discard your session