The Magic of Shinkai Tools: A Beginner’s Guide

Imagine a digital assistant that automatically answers customer emails using your business's key information stored in documents. Or a tool that efficiently manages a cryptocurrency wallet by analyzing the Google Trends lifecycle of specific coins. This is the kind of magic Shinkai Tools brings to life—and you don’t need to be a tech wizard to appreciate it. In this article, we’ll peel back the curtain to explore how these tools work, what they’re made of, and why they’re so powerful.

Disclaimer: This is part 1 of a 3-part series exploring Shinkai Tools. In this article, we’ll cover what Shinkai Tools are and what they’re made of. Part 2 will dive into how to use and import these tools inside the app, while Part 3 will focus on effectively building your own tools.

What Makes a Tool, Well, a Tool?

At its core, a Shinkai Tool is like a digital utility: you provide specific inputs, and it delivers precise outputs. Think of it as your personal problem-solver. To make it relatable, let’s use a simple example: buying ice cream.

The Ice Cream Example 🍦

Meet Mr. Tool. His job? Fetch ice cream for you in his car. To do this, you’ll provide:

1. Money (the budget for the ice cream).
2. The shop’s address (so he knows where to go).

Your instruction might sound like: “Here’s $10 and the address; please get as much ice cream as possible.” Mr. Tool then processes these inputs and delivers the output: your ice cream and any leftover change.

Shinkai Tools operate in a similar way, standardizing inputs and outputs to work seamlessly within automated systems, often with minimal human intervention. This makes them perfect for integration with large language models (LLMs).

Configuration: The Key to Consistency

Much like Mr. Tool’s calm demeanor and dependable car, every Shinkai Tool has a configuration that ensures it performs consistently. This configuration might include:

• API keys or credentials for accessing external systems.
• Operational parameters that define behavior and interpret inputs.

For example, a tool designed to answer emails might include a configuration that specifies its tone (professional, friendly, etc.). Some tools require little to no setup, while others are fine-tuned for specific tasks.

Config vs. Inputs: What’s the Difference?

It’s easy to confuse configuration with inputs, but they serve distinct purposes:

• Configuration: Sets the tool’s environment and behavior. Think of it as the tool’s personality and resources—the API keys, tone settings, or access permissions it needs to function.
• Inputs: These are the specific pieces of data you provide for each task. For example, an email reply tool’s input might be the email text and subject line.

In short, the configuration defines how the tool operates, while the inputs specify what the tool operates on. Keeping this distinction clear ensures your tools function smoothly and predictably.

The Building Blocks

Every Shinkai Tool consists of:

1. Code: The brain, written in Python or TypeScript, that handles tasks and processes inputs.
2. Metadata: A JSON file outlining inputs, outputs, and configurations. This makes tools easy to understand and use, both for humans and LLMs.

These components work together to make tools predictable and reusable, streamlining the creation of complex systems.

Tools Within Tools

Why start from scratch when you can reuse existing tools? For instance, if you’ve already built an ice cream-fetching tool and a banana-fetching tool, you can integrate them into a new tool for making banana splits. This modular approach saves time and fosters innovation.

For example, the email answering tool integrates several smaller tools to handle its workflow. These include an email fetcher to retrieve messages from a server, an email sender to deliver responses, an SQL query tool for storing processed emails and preventing duplicate replies, and an LLM prompt processor to generate the response text automatically.

At Shinkai, we strive to provide a comprehensive suite of foundational tools that can act as the building blocks for more intricate systems. With pre-configured solutions for tasks like web searches, database and memory management, and email handling, we aim to relieve tool builders—whether human or AI—of the need to reinvent common infrastructure. This way, they can channel their energy into crafting creative and innovative solutions.

Wrapping Up

Shinkai Tools are all about empowering creativity while simplifying complexity. Whether you’re crafting a simple utility or building a sophisticated system, these tools provide the foundation to bring your ideas to life.
In our next article, we’ll showcase real-world examples of Shinkai Tools in action, from importing and exporting to integration with LLMs. We’ll also explore how tools can be used within a Shinkai chat instance to handle dynamic tasks or configured for scheduled operations that run periodically, like automating routine updates or data analysis.

But first, it’s time for a well-earned break. I’ll be enjoying some ice cream while dreaming up the next tool idea. What will you create next?