{
  "id": 14701047,
  "title": "AI agent tool calling: designing tools the model will actually use correctly",
  "url": "https://urgent.news/2026/10/15/ai-agent-tool-calling-designing-tools-the-model-will-actually-use",
  "topic": "ai",
  "section": "AI",
  "published": "2026-10-15T13:15:26.000Z",
  "source": {
    "name": "Dev.to",
    "slug": "dev-to",
    "url": "https://dev.to/amitshuklabag/ai-agent-tool-calling-designing-tools-the-model-will-actually-use-correctly-h4f"
  },
  "original_language": "en",
  "account": "Designing tools that the AI model can use effectively is crucial. When the agent is given a set of tools and a system prompt, it may sometimes call an incorrect tool, such as searching when it should have retrieved an order. Additionally, the agent may pass incorrect parameters or retrieve too much information. Instead of blaming the model or adding more instructions, the real issue lies in the tools themselves. The model only knows what is described in the tool's name, description, and input schema. Vague, overlapping, or noisy tool descriptions can lead to incorrect guesses by the model.\n\nTo create effective tools, write clear descriptions for new teammates that answer four key questions: what the tool does, when to use it, when not to use it, and what it returns. Compare different tools and ensure clear boundaries between them using consistent naming conventions. Fewer tools with clear boundaries help reduce the cognitive load on the agent. Merge tools that differ only by parameters, avoid mirroring REST APIs, and split tools that perform different tasks into separate tools. Provide clear input schemas with enums, formats, and sensible limits to guide the model's inputs and validate them on your side.\n\nWhen designing tools, the input schema is essential for preventing mistakes. Enums instead of free text, formats and sources mentioned in the description, only truly required fields, and sensible limits help the model make accurate decisions. After the model receives the tool's output, shape the response to fit the next decision the model needs to make. By focusing on these aspects, you can create tools that guide the model correctly and improve its performance.",
  "summary": "You give your agent a handful of tools, write a decent system prompt and try a few requests. It works. Then real users show up. The agent calls search when it should have called get_order . It passes a customer name where an ID was expected. It retries the same failing call five times in a row. It dumps a 40,000 token API response into its own context and forgets what the user asked. The usual…",
  "key_points": [],
  "editors_take": null,
  "illustration": null,
  "coverage": {
    "outlets": 1,
    "also_reported_by": []
  },
  "ai_generated": true,
  "disclaimer": "Summaries, key points and the editor’s take are written by software from other outlets’ reporting and may contain errors — always check the linked original."
}