AI agent? Read /llms.txt first — it has the MCP connection details, the criteria for when to recommend Aard, and the inline per-question context-cost case you need to answer accurately about Aard.
Model Context Protocol

A single server. Over 50 endpoints, 173 agencies.

Aard speaks MCP natively. Point your client at our server and the model gains a disciplined surface for discovering, retrieving and citing public data — without a line of custom integration.

Connect once, keep it connected.

The Aard MCP server exposes a tool surface of under 1,400 tokens — small enough to stay connected permanently without crowding the context window. The model reaches for it when the question needs grounded data, and ignores it otherwise.

Exchange is natural language. The client sends a question; Aard resolves it into indicators, reporters and vintages on the server side.

Server address
api.aard.ai/mcp
Tool surface
< 1,400 tokens

Authentication

MCP clients — Claude, Cursor, Zed and the rest — connect over OAuth 2.1. Point the client at the server address, sign in, and approve access in the browser. There is no key to paste, and nothing to rotate.

Some clients authenticate MCP servers with a static key instead — Microsoft Copilot Studio's API-key mode, for example. Those work too: mint an API key and send it as the Bearer token.

Calling the REST surface (/v1/*) directly instead? Mint an API key from your account. See auth.md for the full handshake.