Use with AI agents

The Data API is also an MCP (Model Context Protocol) server. Every endpoint is a tool named after its operation, such as getAsset or listAssets, and every tool call is an ordinary API call: the same key, rate limit and credits, and it shows on your usage page like any other call.

POSThttps://api.arzdigital.com/mcp

1. Get a key

Use a key from the keys page. MCP clients send it as "Authorization: Bearer YOUR_API_KEY"; the server passes it on to the API. Paste it below to fill in the examples. It stays in this browser tab.

Paste a key to fill in the examples and try calls. It stays in this tab until you reload. It is never saved. Get a key

Claude Code

Run this once in a terminal. Add --scope user to use the server in every project.

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

Then run /mcp in Claude Code: arzdigital should be connected, with one tool per endpoint.

Cursor

Add the server to .cursor/mcp.json in your project, or to ~/.cursor/mcp.json for every project.

.cursor/mcp.json
{
  "mcpServers": {
    "arzdigital": {
      "url": "https://api.arzdigital.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_API_KEY"
      }
    }
  }
}

Claude API (MCP connector)

Claude connects to the server for you during the request. Give it the server in mcp_servers and turn its tools on with an mcp_toolset entry.

Python
import anthropic

client = anthropic.Anthropic()

response = client.beta.messages.create(
    model="claude-opus-5-5",
    max_tokens=4096,
    betas=["mcp-client-2025-11-20"],
    mcp_servers=[{
        "type": "url",
        "url": "https://api.arzdigital.com/mcp",
        "name": "arzdigital",
        "authorization_token": "YOUR_API_KEY",
    }],
    tools=[{"type": "mcp_toolset", "mcp_server_name": "arzdigital"}],
    messages=[{"role": "user", "content": "What are Bitcoin's price and market cap right now?"}],
)
print(response.content)

The key goes in authorization_token. Anthropic's servers make the calls, so keep this code on your server, not in a browser.

What a tool call returns

Two text blocks: the JSON body exactly as the endpoint returns it, then a line with the credits the call cost and what is left. An error comes back as a tool error with the API's error code, and is not charged.

getAsset · slug=bitcoin
{"status":"success","data":{"slug":"bitcoin","symbol":"BTC", …}}

Credits charged: 1. Credits remaining: 9999. Requests left this minute: 29 of 30.

Cost and context size

  • A tool costs what its endpoint costs. Each tool's description states the price, so the agent can see it before calling.
  • getKeyInfo is free. It shows the plan, balance and limits.
  • Everything a tool returns goes into the model's context. Ask for small pages with limit, and for short windows on charts. There is no field selection yet: a call returns the endpoint's full response.

Not supported yet

  • Sign-in with OAuth. Connectors in claude.ai and ChatGPT need it, so for now use a client that can send a header: Claude Code, Cursor, the Claude API, or your own agent.
  • Server-sent events. The server answers each request with plain JSON, which every current MCP client supports.