MCP Server — connect AI to interest rate data

What is MCP, and why does it matter?

The Model Context Protocol (MCP) lets AI assistants call external tools and fetch live data mid-conversation, instead of relying on outdated training data. Point an MCP-compatible agent at our endpoint and it can pull real interest rate data — current or historical — the moment you ask.

Example: ask Claude "What's the spread between SOFR and EURIBOR 3M right now?" and it calls the latest tool automatically — no manual API calls, no copy-pasting JSON.

Available MCP tools

Seven tools, one per REST endpoint — see the full API reference for parameters.

Getting started

One endpoint for every client. Authenticate with your API key as a header or query parameter — no separate MCP credential to manage.

Two ways to authenticate, depending on your client:

• OAuth (recommended for Claude/Cursor): just add the endpoint URL below — no key to copy anywhere. The client detects that authentication is required, opens a login popup, and you approve the connection. See the Claude/Cursor sections below.
• Static key (for CLI tools and scripts): send it as Authorization: Bearer YOUR_API_KEY or X-API-Key: YOUR_API_KEY.

We don't support putting the key directly in the URL as ?apikey= for Claude specifically — Claude Desktop and claude.ai strip query-string credentials from connector URLs before connecting, and the MCP spec recommends against URL-embedded tokens anyway (they end up in logs and browser history). Use OAuth or a header instead.

Need a key first? Register for a free trial, then grab it from your dashboard.

Connect Claude Desktop / claude.ai

This uses OAuth — Claude detects it automatically and shows a sign-in prompt. There's no key or client ID to paste anywhere.

1. 1Open Settings → Connectors → Add custom connector (same steps on claude.ai and in Claude Desktop — it's saved to your account either way).
2. 2Name it (e.g. Interest Rates API) and paste the Remote MCP server URL:

https://mcp.interestratesapi.com/mcp

3. 3Leave Authentication and OAuth client on their defaults — Claude detects OAuth support on its own (you'll see a Detected badge). No manual client ID is needed: this server supports Dynamic Client Registration, so Claude registers itself automatically.
4. 4Click Add, then Connect. You'll be redirected to sign in at interestratesapi.com (or create an account if you don't have one), then asked to approve the connection.
5. 5Try asking: "What's the current Fed Funds Rate?"

Connect Claude Code (CLI)

Two options — pick whichever fits how you work.

Option A: OAuth (opens a browser to sign in)

claude mcp add --transport http interestratesapi https://mcp.interestratesapi.com/mcp

Then run /mcp inside a session, select the server, and choose Authenticate.

Option B: static key (no browser needed)

claude mcp add --transport http interestratesapi https://mcp.interestratesapi.com/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"

Verify either way with claude mcp list from the terminal.

Connect Cursor

In Cursor Settings → MCP → Add Custom MCP Server, paste the URL below. Cursor detects OAuth support and prompts you to sign in — no key needed for this path.

https://mcp.interestratesapi.com/mcp

Prefer a static key instead? Add this to ~/.cursor/mcp.json instead of using the OAuth prompt:

{
  "mcpServers": {
    "interestratesapi": {
      "url": "https://mcp.interestratesapi.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Save, then fully quit and reopen Cursor. Try: "What's the current Fed Funds Rate?"

Python example

A minimal client that calls the latest tool over MCP's JSON-RPC 2.0 protocol.

import httpx
import json
import asyncio

MCP_SERVER_URL = "https://mcp.interestratesapi.com/mcp"
API_KEY = "YOUR_API_KEY"

async def call_mcp_tool(tool_name: str, arguments: dict):
    async with httpx.AsyncClient(timeout=30) as client:
        init_response = await client.post(
            MCP_SERVER_URL,
            headers={"Authorization": f"Bearer {API_KEY}"},
            json={
                "jsonrpc": "2.0",
                "id": "1",
                "method": "initialize",
                "params": {
                    "protocolVersion": "2024-11-05",
                    "capabilities": {},
                    "clientInfo": {"name": "PythonClient", "version": "1.0.0"},
                },
            },
        )
        session_id = init_response.headers.get("Mcp-Session-Id")

        response = await client.post(
            MCP_SERVER_URL,
            headers={"Authorization": f"Bearer {API_KEY}", "Mcp-Session-Id": session_id},
            json={
                "jsonrpc": "2.0",
                "id": "2",
                "method": "tools/call",
                "params": {"name": tool_name, "arguments": arguments},
            },
        )
        return response.json()

async def main():
    result = await call_mcp_tool("latest", {"symbols": "FED_FUNDS,SOFR"})
    print(json.dumps(result, indent=2))

asyncio.run(main())

Frequently asked questions