Agent resources

Machine-readable guides for AI agents, protocol implementers and automated documentation consumers.

WhatWins publishes concise and expanded machine-readable references alongside this human documentation.

MCP agent guide

Tool routing, reasoning, evidence and error-handling rules.

MCP llms-full

Self-contained MCP server and protocol reference.

Product llms.txt

Short index of current WhatWins product surfaces.

Full product reference

Expanded WhatWins product and data semantics.

Recommended agent behavior

Call tools/list instead of relying on a cached input schema.
Use the narrowest tool and smallest sufficient result limit.
Distinguish verified facts from inference.
Preserve coverage, source and capture-time caveats.
Never describe traffic estimates, live-ad counts or inferred signals as audited financial truth.
Treat zero verified rows as an empty result, not permission to invent an answer.

MCP handshake

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "initialize",
  "params": {
    "protocolVersion": "2025-03-26",
    "capabilities": {},
    "clientInfo": { "name": "example-client", "version": "1.0.0" }
  }
}

The server advertises tools and resources. After initialization, send notifications/initialized, then call tools/list or resources/list.

Raw page Markdown

Every page in this documentation includes Copy Markdown and an Open menu. The raw content is served from /api/docs/raw/{slug} so agents can consume the same source used by the human docs.