> ## Documentation Index
> Fetch the complete documentation index at: https://docs.devic.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Integration with LLMs

> Connect your MCPs with language models like ChatGPT, Claude, or Cursor, and enable their use from agents and assistants in Devic.

## Introduction

Integrating with **language models (LLMs)** allows your tools and services (MCPs) to be consumed directly by conversational models such as **ChatGPT**, **Claude**, **Cursor**, or even by **Agents** and **Assistants** defined within the Devic platform.

This turns your MCPs into **live conversational extensions**, capable of executing real functions from your SaaS or infrastructure when a model invokes them through natural language.

***

## Example: integration with ChatGPT

When a custom MCP is published through its URL, **it can be made available to any language model** compatible with the MCP standard.\
This includes ChatGPT, which allows you to configure custom connectors and invoke tools directly through conversation.

<img src="https://mintcdn.com/devic/7qACVhW-UXP_jBFt/chatgpt-configuration.png?fit=max&auto=format&n=7qACVhW-UXP_jBFt&q=85&s=1a8b2c16861db85f585712b63ca67894" alt="Example of MCP integration with ChatGPT – connector configuration" width="1915" height="983" data-path="chatgpt-configuration.png" />

### How to enable it

1. Go to the **Connectors** section in ChatGPT.
2. Enable **Developer Mode**.
3. Create a new custom connector by entering the **MCP URL** (e.g., `https://mcp.devic.ai/suntropy-solar-service`).
4. Select the **actions (tools)** you want to expose.
5. Save the changes — ChatGPT will now be able to communicate with your MCP.

<img src="https://mintcdn.com/devic/7qACVhW-UXP_jBFt/create-new-mcp.png?fit=max&auto=format&n=7qACVhW-UXP_jBFt&q=85&s=08ae60dc015bcd77e793f81e810791cb" alt="Example of ChatGPT Developer Mode with MCP connected" width="1915" height="983" data-path="create-new-mcp.png" />

> 💡 **Tip:** From this point on, ChatGPT can invoke your tools directly within the conversational flow, allowing users to interact with your SaaS via natural language.

***

## Example of real execution

Once the MCP is connected, users can ask natural-language questions such as:

> *“How many solar inverters do I have available?”*\
> *“Show me the summary of solar studies for September.”*

The model, through the MCP, executes the corresponding tools and renders the visual results defined with widgets.

<img src="https://mintcdn.com/devic/7qACVhW-UXP_jBFt/integration-example.png?fit=max&auto=format&n=7qACVhW-UXP_jBFt&q=85&s=af75f048e7d260c560482d11a2272b18" alt="Example of widget rendered inside ChatGPT with real-time data" width="1256" height="792" data-path="integration-example.png" />

This not only displays data but can do so with **your SaaS branding**, thanks to widgets defined in Devic.

***

## Integration from Agents and Assistants

Beyond integration with ChatGPT or Claude, **MCPs can also be consumed directly from Agents and Assistants configured in Devic**.\
This allows your internal automations, workflows, or enterprise assistants to benefit from the same integration capabilities as an external language model.

* **Agents** can invoke MCPs within their task flows or scheduled executions.\
  See the specific documentation:\
  👉 [Agents → Tools](../agents/tools)

* **Assistants** can use MCPs during conversations or project integrations, showing interactive results or embedding widgets.\
  See the specific documentation:\
  👉 [Assistants → Tools](../assistants/tools)

<img src="https://mintcdn.com/devic/7qACVhW-UXP_jBFt/add-my-mcp-agent.png?fit=max&auto=format&n=7qACVhW-UXP_jBFt&q=85&s=bbcf52b73f1754dbbc53cf9ee449fcf2" alt="Tool selection view from an agent in Devic" width="1915" height="983" data-path="add-my-mcp-agent.png" />

##

<img src="https://mintcdn.com/devic/7qACVhW-UXP_jBFt/add-my-mcp2.png?fit=max&auto=format&n=7qACVhW-UXP_jBFt&q=85&s=b6b274c236f3317886690f7bb1179fd4" alt="Assistant configuring MCP tools in Devic" width="1915" height="983" data-path="add-my-mcp2.png" />

***

## Next steps

<CardGroup cols={2}>
  <Card title="Agents" icon="robot" href="/devic/agents">
    Learn how Devic agents can execute autonomous processes, make data-driven decisions, and orchestrate intelligent tools within your workflows.
  </Card>

  <Card title="Assistants" icon="user" href="/devic/assistants">
    Discover how assistants enable real-time conversational interactions, integrating data and tools from your SaaS environment.
  </Card>

  <Card title="Databases" icon="database" href="/devic/databases">
    Explore how to create, query, and connect internal or vector databases to enhance agents’ and assistants’ reasoning capabilities.
  </Card>

  <Card title="Built-in Tools" icon="wrench" href="/devic/mcps/built-in-tools">
    Browse the complete list of integrated tools available in Devic and learn how to use them within your MCPs or custom agents.
  </Card>
</CardGroup>