HashiConf 2025 Don't miss the live stream of HashiConf Day 2 happening now View live stream
Terraform Home

Prompt a model connected to the Terraform MCP server

This topic provides example prompts for helping you write Terraform configuration after installing the Terraform model context protocol (MCP) server. Refer to Deploy the Terraform MCP server for instructions on how to deploy the server.

Overview

To trigger the MCP server tools, ask your LLM questions about your provider configuration or general questions about Terraform providers and modules. Note that for your LLM to use the server, your questions must pertain to provider configuration from the registry. The LLM uses other sources, such as content from the internet, to respond to general questions about Terraform configuration or requests to generate code.

Your client may use specific characters to invoke tools from connected MCP servers. For example, use #terraform in the Visual Studio Code chat interface to direct your connected model to use the Terraform MCP server. Refer to the documentation for your client or model for instructions on how to trigger specific behaviors.

Requirements

Verify that you have deployed the Terraform MCP server and configured your client. Refer to Deploy the Terraform MCP server for instructions.

Prompt your MCP client

Your model may still decide to use other resources to formulate responses, but you can use the following guidelines to help you create prompts that are more likely to be routed through the server:

  • Include the names of specific providers or provider modules. For example, use google_compute_disk in your prompt instead of the compute disk.
  • Use the full Terraform resource address in your query. Refer to Resource Address Reference in the Terraform CLI documentation for additional information about resource addresses.
  • Ask for the latest information about the provider or module.

Examples

Use the following set of prompts as a guide to help you write your own configuration. The responses are returned from the server using ChatGPT 4.1 in GitHub Copilot on VS Code. Your responses may vary.

These examples show the prompt, the server's API request to the Terraform registry, and a representative AI response. Review responses carefully before applying any configuration. The server retrieves current registry documentation, but AI models can still misinterpret or omit argument details.

Google provider AI resources

The server obtains a list of Google provider document IDs so that you can decide which resources to create.

Prompt

I need help understanding what resources are available 
in the Google provider that are for AI

Request

The server sends the following request payload to the registry: 

{
  "provider_Data_Type": "resources",
  "provider_Name": "google",
  "provider_Namespace": "hashicorp",
  "providerVersion": "latest",
  "serviceSlug": "ai"
}

What the server does

The server calls search_providers with serviceSlug: "ai" to filter the Google provider's documentation index to AI-related entries and returns the matching document IDs and titles. The providerVersion: "latest" field ensures the results reflect the most recently published version of the provider.

Response

The Google Terraform provider offers several resources related to AI, primarily under the Vertex AI and Document AI services. Here are the main AI-related resources available:
 
**Vertex AI Resources**
 
- `vertex_ai_dataset`
- `vertex_ai_deployment_resource_pool`
- `vertex_ai_endpoint`
- `vertex_ai_endpoint_iam`
- `vertex_ai_feature_group`
- `vertex_ai_featurestore`
- `vertex_ai_index`
- `vertex_ai_metadata_store`
- `vertex_ai_tensorboard`
- `document_ai_processor`
- `document_ai_warehouse_document_schema`
 
These resources allow you to manage datasets, endpoints, feature stores, indexes, metadata, and more for AI/ML workflows on Google Cloud, as well as Document AI processors and schemas.

What to verify

The resource names in the response, such as vertex_ai_dataset and vertex_ai_endpoint, are derived from current registry documentation. Confirm that a resource exists in your target provider version before writing configuration. You can check the Google provider resource list directly.

What to do next

To retrieve the full argument schema and an example configuration for any of these resources, follow up with a prompt such as:

Show me the configuration schema for google_vertex_ai_endpoint

Azure storage buckets

The following prompt gets a list of provider documentation IDs. Then, it calls the appropriate documentation for the use case.

Prompt

I need help setting up storage buckets in the azure provider

Request

The server sends the following request payloads to the registry:

{
  "providerDataType": "resources",
  "providerName": "azurerm",
  "providerNamespace": "hashicorp",
  "providerVersion": "latest",
  "serviceSlug": "storage_account"
}

{
  "providerDocID": "8957547"
}

{
  "providerDocID": "8957555"
}

What the server does

The server makes two sequential calls:

  1. search_providers with serviceSlug: "storage_account" to find documentation IDs for storage-related resources in the azurerm provider
  2. get_provider_details for each returned document ID to retrieve the full documentation, including example HCL configurations

This two-step pattern of search then fetch is how the server resolves broad questions into specific, versioned documentation.

Response

The server retrieves several responses from each tool, including example configurations from the registry. This response is one of the examples.

Available Documentation (top matches) for resources in Terraform provider hashicorp/azurerm version: 4.29.0
 
# azurerm_storage_account
 
Manages an Azure Storage Account.
 
## Example Usage
 
```hcl
resource "azurerm_resource_group" "example" {
  name     = "example-resources"
  location = "West Europe"
}
 
resource "azurerm_storage_account" "example" {
  name                     = "storageaccountname"
  resource_group_name      = azurerm_resource_group.example.name
  location                 = azurerm_resource_group.example.location
  account_tier             = "Standard"
  account_replication_type = "GRS"
 
  tags = {
    environment = "staging"
  }
}
```
 
To set up storage buckets in Azure using the Terraform azurerm provider, you need to use the azurerm_storage_account resource. Here's a quick guide and example configuration:
 
1. Terraform Resource for Azure Storage Account
 
```hcl
 resource "azurerm_resource_group" "example" {
  name     = "example-resources"
  location = "West Europe"
}
 
resource "azurerm_storage_account" "example" {
  name                     = "storageaccountname" # must be globally unique, lowercase
  resource_group_name      = azurerm_resource_group.example.name
  location                 = azurerm_resource_group.example.location
  account_tier             = "Standard"           # or "Premium"
  account_replication_type = "GRS"                # or "LRS", "RAGRS", "ZRS", etc.
 
  tags = {
   environment = "staging"
  }
}
```

What to verify

The HCL in the response is pulled from the live registry for azurerm version 4.29.0. The name value for azurerm_storage_account must be globally unique and lowercase. Confirm the account_replication_type value is valid for your Azure region before applying. Not all replication types are available in all regions. Refer to the azurerm_storage_account documentation for the full argument reference.

What to do next

To add a blob container to the storage account, follow up with:

How do I add an azurerm_storage_container to this storage account?

Create an IBM VM

The following prompt gets a list of modules:

Is there a module to help with setting up an IBM VM?

Request

The server sends the following request payload: 

{
  "moduleQuery": "ibm vm"
}

What the server does

The server calls search_modules with the query "ibm vm" to search the public Terraform registry for modules matching that text. It returns module metadata including the namespace, module ID, version, and source repository. The response does not retrieve the full module documentation because it is a discovery step.

Response

The client prints the following response using the context acquired from the MCP server.

Yes, there is a Terraform module to help with setting up an IBM VM:
 
- IBM VM Module
- ID: intel/ibm-vm/intel/1.2.2
- Description: Intel Cloud Optimization Module - IBM VM
- Module Version: 1.2.2
- Namespace: intel
- Source: GitHub - intel/terraform-intel-ibm-vm
 
You can use this module to set up and manage IBM VMs with Terraform.
 
If you need more details or usage examples, let me know.

What to verify

The module ID intel/ibm-vm/intel and version 1.2.2 were current at the time this example was captured. Verify the latest version and review the required inputs before using the module. You can inspect the module at registry.terraform.io/modules/intel/ibm-vm/intel.

What to do next

To retrieve the full module documentation, including required inputs, optional variables, and usage examples, follow up with:

Show me the full details for the intel/ibm-vm/intel module
Edit this page on GitHub