fintual_mcp_server

This server enables LLMs to interact with Fintual's API.

Swagger: https://fintual.cl/api-docs/index.html

• Python Server
• Typescript Server (coming soon)

1- Clone this repository

git clone https://github.com/davila7/fintual_mcp_server.git

2- Install uv

curl -LsSf https://astral.sh/uv/install.sh | sh

Go to the fintual_python_server folder and run the following command:

cd fintual_python_server

Run the following command:

uv run mcp install server.py

or Add this to your .codegpt/mcp_config.json for CodeGPT or /path/to/Claude/claude_desktop_config.json for Claude Desktop:

{
  "mcpServers": {
    "Fintual MCP Server": {
      "command": "/path_to_your_uv/.local/bin/uv",
      "args": [
        "run",
        "--with",
        "mcp[cli]",
        "mcp",
        "run",
        "<path to mcp-servers>/fintual_mcp_server/fintual_python_server/server.py"
      ]
    }
  }
}

• Replace /path_to_your_uv/.local/bin/uv with the actual path to your uv executable.
• Replace <path to mcp-servers>/fintual_mcp_server/fintual_python_server/server.py with the actual path to the server.py file of the Fintual MCP server.

The common JSON configuration should be placed in the following file, depending on the tool you are using:

• Claude Desktop: claude_desktop_config.json
• Cursor: .cursor/mcp.jso
• CodeGPT: ~/.codegpt/mcp_config.json

• asset_provider: Get the list of asset providers.
• get_asset_provider_by_id: Get the asset provider by id.
• get_conceptual_assets: Get the list of conceptual assets.
• banks: Get the list of banks.

• assets_provider_prompt: Prompt to use the assets_provider tool.
• conceptual_assets_prompt: Prompt to use the conceptual_assets tool.

1. Question Example: Zurich es una administradora de activos de Fintual ?

2. Question Example: Cuales son los bancos integrados en Fintual ?

3. Question Example: Cuales son los activos conceptuales de BANCHILE?

In Claude Desktop, you can add an speific prompt to use the tools.

• Select assets_provider_prompt

• Add the assets_provider

How it works

• The model will use the tools to answer the questions, but first it will ask for confirmation to use the tools.

• Then the model will read the API response and answer the question.

What's next

• Add more tools to the MCP Server.
• Add more prompts to use the tools.
• Add typescript support to the MCP Server.