terraform MCP Server

Use This MCP server To

Look up every argument a cloud resource accepts Find a well-maintained module for something I need See an example of how a resource is normally written List the data sources a provider offers Check what a provider-specific function does

README

Terraform Registry MCP Server

A Model Context Protocol (MCP) server that provides tools for interacting with the Terraform Registry API. This server enables AI agents to query provider information, resource details, and module metadata.

Important

This project was used as a PoC for a new official Terraform MCP server. This repo has been archived in favor of that one.

Installation

Installing in Cursor

To install and use this MCP server in Cursor:

In Cursor, open Settings (⌘+,) and navigate to the "MCP" tab.
Click "+ Add new MCP server."
Enter the following:
- Name: terraform-registry
- Type: command
- Command: npx -y terraform-mcp-server
Click "Add" then scroll to the server and click "Disabled" to enable the server.
Restart Cursor, if needed, to ensure the MCP server is properly loaded.

terraform-registry MCP settings for Cursor

Installing in Claude Desktop

To install and use this MCP server in Claude Desktop:

In Claude Desktop, open Settings (⌘+,) and navigate to the "Developer" tab.
Click "Edit Config" at the bottom of the window.
Edit the file (~/Library/Application Support/Claude/claude_desktop_config.json) to add the following code, then Save the file.

{
  "mcpServers": {
    "terraform-registry": {
      "command": "npx",
      "args": ["-y", "terraform-mcp-server"]
    }
  }
}

Restart Claude Desktop to ensure the MCP server is properly loaded.

Tools

The following tools are available in this MCP server:

Core Registry Tools

Tool	Description
`providerDetails`	Gets detailed information about a Terraform provider
`resourceUsage`	Gets example usage of a Terraform resource and related resources
`moduleSearch`	Searches for and recommends Terraform modules based on a query
`listDataSources`	Lists all available data sources for a provider and their basic details
`resourceArgumentDetails`	Fetches comprehensive details about a resource type's arguments
`moduleDetails`	Retrieves detailed metadata for a Terraform module
`functionDetails`	Gets details about a Terraform provider function
`providerGuides`	Lists and views provider-specific guides and documentation
`policySearch`	Searches for policy libraries in the Terraform Registry
`policyDetails`	Gets detailed information about a specific policy library

Terraform Cloud Tools

These tools require a Terraform Cloud API token (TFC_TOKEN):

Tool	Description
`listOrganizations`	Lists all organizations the authenticated user has access to
`privateModuleSearch`	Searches for private modules in an organization
`privateModuleDetails`	Gets detailed information about a private module
`explorerQuery`	Queries the Terraform Cloud Explorer API to analyze data
`listWorkspaces`	Lists workspaces in an organization
`workspaceDetails`	Gets detailed information about a specific workspace
`lockWorkspace`	Locks a workspace to prevent runs
`unlockWorkspace`	Unlocks a workspace to allow runs
`listRuns`	Lists runs for a workspace
`runDetails`	Gets detailed information about a specific run
`createRun`	Creates a new run for a workspace
`applyRun`	Applies a run that's been planned
`cancelRun`	Cancels a run that's in progress
`listWorkspaceResources`	Lists resources in a workspace

Resources

The MCP server supports the following resource URIs for listing and reading via the resources/* methods:

Resource Type	Example URI(s)	Description
Providers	`terraform:providers`	List all namespaces/providers
	`terraform:provider:<namespace>/<name>`	Get details for a specific provider
Provider Versions	`terraform:provider:<namespace>/<name>/versions`	List available versions for a provider
Provider Resources	`terraform:provider:<namespace>/<name>/resources`	List resources for a provider
	`terraform:resource:<namespace>/<name>/<resource_name>`	Get details for a specific resource type
Provider Data Sources	`terraform:provider:<namespace>/<name>/dataSources`	List data sources for a provider
	`terraform:dataSource:<namespace>/<name>/<data_source_name>`	Get details for a specific data source
Provider Functions	`terraform:provider:<namespace>/<name>/functions`	List functions for a provider
	`terraform:function:<namespace>/<name>/<function_name>`	Get details for a specific function

The server also supports resources/templates/list to provide templates for creating:

terraform:provider
terraform:resource
terraform:dataSource

Prompts

The following prompts are available for generating contextual responses:

Prompt	Description	Required Arguments
`migrate-clouds`	Generate Terraform code to migrate infrastructure between cloud providers	`sourceCloud`, `targetCloud`, `terraformCode`
`generate-resource-skeleton`	Helps users quickly scaffold new Terraform resources with best practices	`resourceType`
`optimize-terraform-module`	Provides actionable recommendations for improving Terraform code	`terraformCode`
`migrate-provider-version`	Assists with provider version upgrades and breaking changes	`providerName`, `currentVersion`, `targetVersion`, `terraformCode` (optional)
`analyze-workspace-runs`	Analyzes recent run failures and provides troubleshooting guidance for Terraform Cloud workspaces	`workspaceId`, `runsToAnalyze` (optional, default: 5)

Known Issues with Prompts

Note: There is a known issue with the getPrompt functionality that can cause server crashes. The server properly registers prompts and can list them, but direct requests using the getPrompt method may cause connectivity issues. This is being investigated and may be related to SDK compatibility or implementation details. Until resolved, use listPrompts to see available prompts but avoid direct getPrompt calls.

Running the Server

The server runs using stdio transport for MCP communication:

npm install
npm start

Configuration with Environment Variables

The server can be configured using environment variables:

Environment Variable	Description	Default Value
`TERRAFORM_REGISTRY_URL`	Base URL for Terraform Registry API	https://registry.terraform.io
`DEFAULT_PROVIDER_NAMESPACE`	Default namespace for providers	hashicorp
`LOG_LEVEL`	Logging level (error, warn, info, debug)	info
`REQUEST_TIMEOUT_MS`	Timeout for API requests in milliseconds	10000
`RATE_LIMIT_ENABLED`	Enable rate limiting for API requests	false
`RATE_LIMIT_REQUESTS`	Number of requests allowed in time window	60
`RATE_LIMIT_WINDOW_MS`	Time window for rate limiting in milliseconds	60000
`TFC_TOKEN`	Terraform Cloud API token for private registry access (optional)

Example usage with environment variables:

# Set environment variables
export LOG_LEVEL="debug"
export REQUEST_TIMEOUT_MS="15000"
export TFC_TOKEN="your-terraform-cloud-token"

# Run the server
npm start

Testing

See the TESTS.md file for information about testing this project.

terraform-mcp-server FAQ

Is this still maintained?

No. The author archived it in favour of HashiCorp's official Terraform MCP server, and describes this one as the proof of concept that led to it. Nothing has changed here since May 2025.

Should I install it anyway?

Only if you have a specific reason. An archived project gets no fixes, and the official replacement it names is the safer choice.

Did it need an account or a key?

Not for registry lookups — those are public. A Terraform Cloud token was optional, and supplying one unlocked a second set of tools that reach your own workspaces.

Could it change my real infrastructure?

Yes, with a Terraform Cloud token. It could start a run, mark it as a teardown, turn on automatic approval and approve a run, with nothing asking you to confirm first. Without a token it is lookup only.

Which apps did it work in?

Cursor and Claude Desktop both had documented setups, and it also shipped as a container image.

How many tools were there?

Ten without a Terraform Cloud token and twenty-four with one.

Can I use this to learn how to write Terraform?

It was built for exactly that, but for current documentation you are better off with the official replacement.

Is the licence a problem at work?

No. It is MIT licensed, so commercial use is allowed.