ISO 20022 MCP Suite: documentation.

Cursor

Create .cursor/mcp.json in your project, or ~/.cursor/mcp.json to make the suite available everywhere:

{
  "mcpServers": {
    "iso20022": {
      "command": "uvx",
      "args": ["--from", "iso20022-mcp[all]", "iso20022-mcp"]
    }
  }
}

Windsurf

Same mcpServers shape, in ~/.codeium/windsurf/mcp_config.json. Add the identical iso20022 entry shown for Cursor.

VS Code with GitHub Copilot

Create .vscode/mcp.json in your workspace. VS Code is the odd one out: the top-level key is servers, not mcpServers:

{
  "servers": {
    "iso20022": {
      "type": "stdio",
      "command": "uvx",
      "args": ["--from", "iso20022-mcp[all]", "iso20022-mcp"]
    }
  }
}

Google Gemini CLI

Add an mcpServers block to ~/.gemini/settings.json (or .gemini/settings.json per project), with the identical iso20022 entry shown for Cursor. Transport is inferred: a command key means stdio.

OpenAI Agents SDK

The Python Agents SDK spawns local stdio servers itself through MCPServerStdio, with the same command and args:

async with MCPServerStdio(
    name="iso20022",
    params={
        "command": "uvx",
        "args": ["--from", "iso20022-mcp[all]", "iso20022-mcp"],
    },
) as server:
    agent = Agent(name="Treasury", mcp_servers=[server])

ChatGPT and the OpenAI Responses API. Remote-first: they connect only to MCP servers reachable over the public internet, via Streamable HTTP or HTTP/SSE and a server_url. They do not run local stdio processes, so pairing them with this suite means hosting it yourself; the verified bridge below is the starting point.

Microsoft Copilot Studio. Adds MCP servers to agents as tools through Power Platform connectors, over the Streamable HTTP transport only. No local stdio support.

Zapier MCP. The inverse case: Zapier hosts a remote MCP endpoint exposing Zapier actions to your AI client via a generated mcp.zapier.com URL. It is a place agents call out to, not a way to run this suite.