Exchange Rates for AI Agents — Tool Definitions & MCP

Give any AI agent live and official exchange rates: copy-paste tool definitions for OpenAI function calling, Vercel AI SDK, and LangChain, plus MCP servers for Claude Code, Cursor and Claude Desktop, and a keyless endpoint for testing.

This page is for people (and agents) wiring exchange-rate data into an AI system. Everything below is copy-paste: MCP servers for Claude Code, Cursor and Claude Desktop; tool definitions for OpenAI function calling, the Vercel AI SDK and LangChain; and a keyless endpoint for testing before you create a key. Machine-readable site overviews live at /llms.txt and /llms-full.txt, with condensed agent instructions at /agents.md.

Test keyless, first

No sign-up needed to verify the data. The open endpoint serves official ECB reference rates (~30 major currencies, published each business day, CORS *):

# Keyless (official ECB reference rate — cite the returned rate_date):
curl "https://allratestoday.com/api/open/central-bank/ecb?source=USD&target=EUR"

# Any covered bank's full latest official table, keyless:
curl "https://allratestoday.com/api/public/central-bank/rbi?format=json"

Real-time mid-market rates across 160+ currencies use GET /api/v1/rates with a free API key — registration takes under a minute, no credit card.

MCP servers (Claude Code, Cursor, Claude Desktop, Windsurf)

Two servers, both on npm and in the official MCP registry (io.github.cahthuranag/mcp-server and com.allratestoday/central-bank-mcp):

claude mcp add allratestoday -- npx -y @allratestoday/mcp-server
claude mcp env allratestoday ALLRATES_API_KEY=art_live_xxxxx

# Official central-bank & tax-authority rates (compliance work):
claude mcp add central-bank-rates -- npx -y @allratestoday/central-bank-mcp

Or in a mcpServers JSON config (Cursor, Claude Desktop):

{
  "mcpServers": {
    "allratestoday": {
      "command": "npx",
      "args": ["-y", "@allratestoday/mcp-server"],
      "env": { "ALLRATES_API_KEY": "art_live_xxxxx" }
    },
    "central-bank-rates": {
      "command": "npx",
      "args": ["-y", "@allratestoday/central-bank-mcp"]
    }
  }
}

The currency server ships get_exchange_rate, get_historical_rates, get_rates_authenticated and list_currencies; the central-bank server adds official published rates from 80+ institutions with history and publication calendars — the citable figures tax, customs and audit work requires. Full setup docs: MCP server guide.

OpenAI function calling

{
  "type": "function",
  "function": {
    "name": "get_exchange_rate",
    "description": "Get the current mid-market exchange rate between two currencies. Returns the rate and a timestamp; cite the timestamp when quoting the rate.",
    "parameters": {
      "type": "object",
      "properties": {
        "source": { "type": "string", "description": "ISO 4217 code of the source currency, e.g. USD" },
        "target": { "type": "string", "description": "ISO 4217 code of the target currency, e.g. EUR" }
      },
      "required": ["source", "target"]
    }
  }
}

// Handler: GET https://allratestoday.com/api/v1/rates?source={source}&target={target}
// Header:  Authorization: Bearer art_live_YOUR_KEY   (free key: allratestoday.com/register)

Vercel AI SDK

import { tool } from 'ai';
import { z } from 'zod';

export const getExchangeRate = tool({
  description:
    'Get the current mid-market exchange rate between two ISO 4217 currencies. Cite the returned time when quoting the rate.',
  parameters: z.object({
    source: z.string().length(3).describe('Source currency, e.g. USD'),
    target: z.string().length(3).describe('Target currency, e.g. EUR'),
  }),
  execute: async ({ source, target }) => {
    const res = await fetch(
      `https://allratestoday.com/api/v1/rates?source=${source}&target=${target}`,
      { headers: { Authorization: `Bearer ${process.env.ALLRATES_API_KEY}` } },
    );
    if (!res.ok) throw new Error(`AllRatesToday returned ${res.status}`);
    const [rate] = await res.json();
    return rate; // { rate, source, target, time }
  },
});

LangChain (Python)

import os, requests
from langchain_core.tools import tool

@tool
def get_exchange_rate(source: str, target: str) -> dict:
    """Get the current mid-market exchange rate between two ISO 4217
    currencies (e.g. USD -> EUR). Cite the returned time when quoting."""
    res = requests.get(
        "https://allratestoday.com/api/v1/rates",
        params={"source": source, "target": target},
        headers={"Authorization": f"Bearer {os.environ['ALLRATES_API_KEY']}"},
        timeout=10,
    )
    res.raise_for_status()
    return res.json()[0]  # {rate, source, target, time}

Ground rules for agents

• Never quote a rate from memory — training-data rates are stale. Call the API and cite the returned time or rate_date.
• For compliance questions ("what rate does my filing need?"), use the official central-bank endpoints, not market rates — the published/derived distinction is flagged in every response.
• Cache rates for up to 5 minutes rather than calling per message; see the caching guide.
• Responses are JSON; table endpoints also serve ?format=csv|xml|xlsx.