DEV Community

Soham Ganatra
Soham Ganatra

Posted on • Originally published at blog.composio.dev on

Making the Most of LLMs with AI Agent Tools

Making the Most of LLMs with AI Agent Tools

AI agents are all the rage. With the ever-improving quality of Large Language Models, the demand for AI automation is also increasing. The Large Language or Multi-modal models are efficient at reasoning, summarizing, general question and answering, etc. The efficient reasoning abilities enable LLMs to analyze complex tasks and break them into smaller sub-tasks.

However, to fully leverage their capabilities in complex automation scenarios, LLMs require the right tools. By equipping them with such tools, these models can intelligently decide which tool to use and when to use it during task execution. To make sure the tasks are executed properly the tools need to be reliable.

Composio is the platform that provides production-ready tool integration with LLM frameworks like LangChain, AutoGen, and CrewAi to build reliable AI agents. Composio’s repertoire has 150+ out-of-the-box tools for apps across the genre like CRM, Productivity, SDE, etc, and it also lets you easily add custom tools.

So, this article will explore AI agents, agent tools, and specifically Composio’s tool integrations.

Learning Objectives

  • Learn about AI agents, their definition, working, and usefulness.
  • Understand what agent tools are.
  • Explore Composio tool integrations to empower agents.
  • Learn about custom tools integration on Composio.

What are AI agents?

We now have a brief idea about agents, let’s dig a bit more. So, the Agents are pieces of software that can dynamically interact with their environment, and the AI in the term “AI Agents” refers to Large Language Models (LLMs) or Large Multi-modal Models (LMMs).

LLMs possess great reasoning ability (thanks to extensive training in reasoning tasks). This enables them to analyze a complex task step-by-step. When the LLMs have access to the right tools, they can break down a problem statement and use the right tools to execute tasks as and when needed. The best example of this would be the ChatGPT app itself. It has access to code interpreters, the Internet, and DallE. Based on the given query, it decides which tool to use. If you ask it to create an image, it will use Dalle, for executing codes, it will choose code interpreter. However, the agents do not always need to be in a chat app. We can use external apps like Discord, Slack, GitHub, etc to trigger an agentic workflow. For instance, an agent with Slack and Notion integration will trigger when a message is sent in the Slack channel. The agent will pick up the task, execute it, write it to a Notion doc, and return a confirmation message in the Slack channel.

So, AI agents are LLMs augmented with tools and goals.

What are Agent tools?

Tools are software that enables the LLMs to execute a given task. They are interfaces for LLMs allowing them to interact with the external environment. Tools provide the agency for LLMs to carry out a given task. It's like painting; you need good brushes, colors, and a canvas to make something great.

With this in mind, Composio offers over 150 agent toolkits, each packed with built-in actions and triggers. It's like having a fully stocked art supply store at your disposal. This ensures that whatever the project, you've got the tools needed to tackle it efficiently.

Composio’s Comprehensive Toolkit

Imagine you're managing multiple projects across different platforms. Normally, this could involve a lot of manual coordination and checking in and might become tedious and time-consuming. That's where Composio's tools come into play. For instance, you could set up an automation that syncs your project tasks between GitHub and a project management tool like Trello or Asana. Whenever an issue is updated in GitHub, it updates your project board.

Or you can use Composio’s Typeform and Google Sheet integrations to automate user feedback collection and update it to Google Sheet for further downstream tasks like CRM integration for lead management, trend analysis, etc.

These are only a few examples. You can integrate multiple Composio tools across the categories with the LLM frameworks of your choice to automate complex workflows.

Check out the official tools supported here in the Tools Catalog.

Making the Most of LLMs with AI Agent Tools

For a detailed look at how Composio operates, read this walk-through guide, where Composio’s Slack and Notion integration is used to create an AI research assistant.

Build an AI Research Assistant Using CrewAI and Composio

Custom AI Tools

Composio also gives developers the freedom to build custom integrations for specific needs. All you need to do is follow a few structured steps using the OpenAPI Spec. Here's how to get started:

  • Create or Obtain OpenAPI Spec : Begin by acquiring the OpenAPI Spec for the application you want to integrate with Composio. If your chosen application lacks an OpenAPI Spec, you can create one using the Swagger Editor.
  • Create the integrations.yaml File : Prepare an integrations.yaml file using the provided base template. This file should be customized to include the authentication schemes suitable for the tool you are integrating, detailing essential aspects such as the application's name, description, and authentication methods.
  • Fill Out Authentication Details :
    • Select the appropriate authentication method—OAuth1, OAuth2, API-KEY, or BASIC—based on what the custom tool supports.
    • For tools like GitHub, utilize OAuth2 and include necessary details such as authorization_url , token_url , default_scopes , and other relevant parameters.
  • Push and Copy Repository URL : Once your integrations.yaml file is ready, push the changes to your repository and copy the URL of this repository.
  • Add Your Custom Tool on Composio :
    • Navigate to the settings page on Composio.
    • Access the "Add Open API spec" section.
    • Upload both the OpenAPI Spec file and the integrations.yaml file.
    • Initiate the integration by clicking on the "Start import" button.
  • Test Your Custom Tool on Composio :
    • Return to the tools catalog on Composio.
    • Locate and select the newly created tool.
    • Connect your account to use the tool and ensure it functions as expected.

With these steps, you can efficiently integrate and automate your workflows using Composio, harnessing the full potential of its expansive toolkit to meet your specific project needs.

Here’s a quick video for adding a custom tool to Composio.

Conclusion

AI Agents are inevitable and in the future, it is safe to assume, that most software systems will integrate AI agents in one way or another, From automating mundane day-to-day tasks to handling complex enterprise operations, the scope is vast. Composo provides a robust platform designed to meet these needs. With 150+ production-ready agent tools with many actions and triggers developers can create reliable and useful AI agents that work.

Tool sets from different categories like Productivity, SDE, CRM, social media, marketing, design, and more can be tailored to specific tasks, ensuring versatility across all business functions. Whether automating communication channels, code deployments, managing customer relationships, enhancing social media engagement, driving marketing campaigns, or facilitating design processes, Composio’s tools adapt to the unique demands of each domain. This will allow businesses to operate more efficiently, reduce costs, and focus human efforts on more creative tasks.

Frequently Asked Question about AI Agent Tools

  1. What are AI agents?

Ans. The AI agents are LLMs augmented with specific tools and goals that enable them to carry out complex tasks autonomously.

  1. What are Agent tools?

Ans. Agent tools are software that serves as the interface between Large Language Models (LLMs) and external applications, enabling the completion of tasks.

  1. What is Composio?

Ans. Composio is a platform that provides production-ready tool integration with LLM frameworks to build reliable AI agents for automating complex workflows.

Top comments (0)