OpenAI

Connect Integrations to
OpenAI

Build advanced AI agents with OpenAI. Connect 600+ integrations, automate workflows, and deploy with ease using Metorial.

Back to OpenAI overview

Handling Tool Calls with OpenAI

When using Metorial with OpenAI, your AI agent can call tools to access integrations. This guide explains how to handle those tool calls.

How Tool Calling Works

  1. You provide available tools to OpenAI
  2. OpenAI decides when to call a tool based on the conversation
  3. You execute the tool call using Metorial
  4. You send the results back to OpenAI
  5. OpenAI uses the results to continue the conversation

Complete Example

import { Metorial } from 'metorial';
import { metorialOpenAI } from '@metorial/openai';
import OpenAI from 'openai';

const metorial = new Metorial({
  apiKey: process.env.METORIAL_API_KEY
});

const openai = new OpenAI({
  apiKey: process.env.OPENAI_API_KEY
});

await metorial.withProviderSession(
  metorialOpenAI.chatCompletions,
  { serverDeployments: ['your-server-deployment-id'] },
  async session => {
    let messages = [
      { role: 'user', content: 'What are my upcoming meetings today?' }
    ];

    // Initial request with tools
    let response = await openai.chat.completions.create({
      model: 'gpt-4o',
      messages,
      tools: session.tools
    });

    // Check if the model wants to call tools
    let toolCalls = response.choices[0]?.message.tool_calls;
    
    if (toolCalls) {
      // Execute the tool calls through Metorial
      let toolResponses = await session.callTools(toolCalls);
      
      // Add assistant's message and tool responses to conversation
      messages.push(
        { role: 'assistant', tool_calls: toolCalls },
        ...toolResponses
      );
      
      // Get final response with tool results
      response = await openai.chat.completions.create({
        model: 'gpt-4o',
        messages,
        tools: session.tools
      });
    }

    console.log(response.choices[0]?.message.content);
  }
);

Key Points

  • session.tools provides the formatted tools to OpenAI
  • session.callTools() executes the tool calls and returns properly formatted responses
  • Always add both the assistant's tool calls and the tool responses to your message history
  • The model may make multiple tool calls in succession

OpenAI on Metorial

Supercharge your OpenAI GPT applications with Metorial's extensive integration library featuring over 600 tools and services. Our MCP-powered platform makes it simple to connect GPT-4, GPT-4o, and other OpenAI models to the APIs your applications need. With Metorial's Python and TypeScript SDKs, adding integrations to your OpenAI-based agents and applications takes just a couple of lines of code—no more spending weeks building custom connectors. Whether you're creating chatbots, assistants, automation workflows, or AI-powered features with OpenAI's industry-leading language models, Metorial handles all the integration complexity. Connect to productivity tools like Google Workspace and Microsoft 365, CRMs like Salesforce and HubSpot, development tools like GitHub and Jira, and hundreds of other popular services instantly. Our open-source approach gives you full transparency and control while eliminating the burden of maintaining integration code, handling authentication, and managing API changes. Focus your team on building unique AI experiences and delivering value to users—let Metorial manage the integration infrastructure that powers your OpenAI applications.

Connect anything. Anywhere.

Ready to build with Metorial?

Let's take your AI-powered applications to the next level, together.

About Metorial

Metorial provides developers with instant access to 600+ MCP servers for building AI agents that can interact with real-world tools and services. Built on MCP, Metorial simplifies agent tool integration by offering pre-configured connections to popular platforms like Google Drive, Slack, GitHub, Notion, and hundreds of other APIs. Our platform supports all major AI agent frameworks—including LangChain, AutoGen, CrewAI, and LangGraph—enabling developers to add tool calling capabilities to their agents in just a few lines of code. By eliminating the need for custom integration code, Metorial helps AI developers move from prototype to production faster while maintaining security and reliability. Whether you're building autonomous research agents, customer service bots, or workflow automation tools, Metorial's MCP server library provides the integrations you need to connect your agents to the real world.

Star us on GitHub