APIs & Agentic AI

MCP tool-definition linter

Lint MCP or function-calling tool definitions for what fails silently — descriptions that don't say when to use a tool, undescribed parameters, missing safety annotations.

A model picks a tool by reading its description, and fills the arguments by reading the parameter descriptions. A definition written for a human — “Gets an order” — gives it nothing to decide when to call it. Nothing errors; the behaviour is just quietly wrong.

MCP tools array, or function-calling definitions

35 lines · 1,013 characters

{
  "tools": [
    {
      "name": "get_order",
      "description": "Gets an order",
      "inputSchema": {
        "type": "object",
        "properties": { "orderId": { "type": "string" } },
        "required": ["orderId"]
      }
    },
    {
      "name": "fetch_order",
      "description": "Gets an order from the system",
      "inputSchema": {
        "type": "object",
        "properties": { "id": { "type": "string", "description": "The order identifier" } },
        "required": ["id"]
      }
    },
    {
      "name": "cancel_order",
      "description": "Cancels an order. Use this only when the user has explicitly confirmed the order number and understands the cancellation is final.",
      "inputSchema": {
        "type": "object",
        "properties": {
          "orderId": { "type": "string", "description": "The order to cancel" },
          "reason": { "type": "string" },
          "status": { "type": "string" }
        },
        "required": ["orderId", "notes"]
      }
    }
  ]
}

Inspection files can contain the original data and identifiers. Review them before sharing; downloads are processed locally. This is a parser inspection, not certification.

Tool definitions are an integration contract wearing new vocabulary — the same authentication, scope, validation and audit questions any interface has answered for twenty years. Why we think the work sits below the model.

Checks the things that fail silently: descriptions that say what a tool is but not when to use it, undescribed parameters that become hallucinated arguments, free-string fields that should be enums, write-like tools with no safety annotations, and near-duplicate descriptions that make selection arbitrary. Runs in your browser.

Stay ahead of the integration layer.

Integronauts Signal — practical enterprise integration, API and agentic AI thinking. The tools stay free either way.

Integronauts Signal is launching soon.

Practical enterprise integration, APIs, agentic AI and emerging architecture. Sign-up opens when the list does — nothing to enter yet.