Integrating LLMs via REST API with TMS AI Studio's TTMSMCPCloudAI Component

The surge in Large Language Models (LLMs) such as ChatGPT, Claude, Mistral, and others has revolutionized how applications interact with language. Whether you want to translate content, summarize data, or build intelligent assistants, these models provide a consistent and scalable interface through REST APIs.

In this article, we’ll explore how you can abstractly and efficiently work with a wide range of LLM services—like OpenAI, DeepSeek, Claude, Gemini, Perplexity, Grok, Ollama, Mistral —using the powerful TTMSMCPCloudAI component from TMS AI Studio.

A Unified Approach to LLM Integration

Each cloud LLM provider offers slightly different API semantics, authentication methods, and payload structures. However, at a higher level, their interfaces boil down to a simple pattern:

1. Send text input ("prompt")

2. Receive generated text response ("completion")

This is commonly referred to as the Completion REST API.

With TTMSMCPCloudAI, you don’t need to worry about low-level details. The component abstracts away REST communication, authentication, and request building, so you can focus purely on integrating language capabilities into your app.

Supported AI services include at this moment:

• OpenAI (ChatGPT)
  Leading AI research and deployment company, creator of GPT-4 and ChatGPT.

• Gemini (Google AI)
  Google's flagship multimodal AI family developed by Google DeepMind.

• Claude (Anthropic)
  Anthropic’s family of helpful, harmless, and honest AI assistants.

• Grok (xAI)
  AI developed by Elon Musk’s xAI, integrated with X (formerly Twitter).

• DeepSeek
  Open-source and commercial LLMs focused on code and reasoning tasks.

• Mistral
  European AI company offering open-weight LLMs like Mistral and Mixtral.

• Ollama
  A local runtime for running LLMs on your own machine with a simple CLI.

• Perplexity AI
  An AI-powered answer engine combining web search and conversational AI.

Common Use Cases

You can use these models for a variety of natural language processing tasks:

• Translation between languages

• Summarization of documents or datasets

• Information retrieval from user input

• Conversational agents and chatbots

• And much more you'll uncover in the next article on function calling! 

All these use cases follow the same underlying pattern: send context → receive a response. This makes them perfect candidates for abstraction.

Step-by-Step Integration Workflow

The typical workflow using the TTMSMCPCloudAI component looks like this:

1. Request an API key from the LLM provider (e.g., OpenAI, Anthropic).

2. Select your model (e.g., gpt-4, claude-3-7-sonnet, mistral-medium).

3. Set the prompt context, and optionally the system role (instructions for the model’s behavior).

4. Send the request via HTTPS POST.

5. Receive and process the generated response text.

All of this is encapsulated cleanly by the component, making it straightforward to switch between providers or models.

Code Example: Ask Claude About Development Tools

Here's a practical example of how to call Anthropic's Claude model using TTMSMCPCloudAI in a Delphi or C++Builder application:

pascal
CopyEdit
TMSMCPCloudAI.Settings.ClaudeModel := 'claude-3-7-sonnet-20250219';
TMSMCPCloudAI.APIKeys.Claude := KEY_CLAUDE;

TMSMCPCloudAI.Service := aiClaude;

TMSMCPCloudAI.Context.Text := 
  'What is a powerful development tool for creating native Windows ' +
  'applications with a RAD component based visual approach?';

TMSMCPCloudAI.SystemRole.Text := 'You are an IT expert';

TMSMCPCloudAI.Execute();

Once Execute is called, the request is sent, and the response text will be available via the result event or property for further use in your application.

And of course it is reassuring to see that the LLM responds with Delphi as its first recommendation 😉:

Why Use TTMSMCPCloudAI?

- Rapid Development: Eliminate the boilerplate of REST clients and JSON handling.
- Flexibility: Switch between LLM providers without rewriting logic.
- RAD Integration: Seamlessly integrates into the Delphi or C++Builder apps built for Windows, macOS, Linux, iOS, Android.
- Cross-platform: Build once, deploy anywhere.

Get started

As LLMs become increasingly critical in building intelligent applications, having a robust and unified abstraction like TTMSMCPCloudAI allows you to stay focused on the logic and user experience—not the plumbing.

In upcoming articles, we’ll dive deeper into function calling, RAG, agents, MCP servers & clients.  
If you have an active TMS ALL-ACCESS license, you have access to the TMS AI Studio that uses the TTMSMCPCloudAI component but also has everything on board to let you build MCP servers and clients.