This is the installation page for the Elastio MCP (Model Context Protocol) server, which provides AI assistants like Claude access to Elastio's cyber resilience platform.
The MCP server requires configuration to connect to your Elastio tenant. There are two required parameters:
Tenant URL: The URL of your Elastio tenant
ELASTIO_TENANT_URL--tenant-urlAPI Key: Your Elastio API access token
ELASTIO_API_KEY--api-keyIn the MCP client examples below, we use environment variables since most MCP hosts natively support them, and it simplifies the configuration JSON. However, you can also pass these parameters are command line arguments, for example:
npx -y https://mcp.elastio.com/elastio-mcp-0.9.0.tgz \
--tenant-url "https://your-tenant.app.elastio.com" \
--api-key "your-access-token"Elastio's MCP server works with any compatible MCP host which includes many AI assistant tools. Below are setup instructions for some popular MCP-compatible platforms. We have provided example configs for each of these tools, but please refer to the latest documentation for your specific AI assistant for more specific instructions and troubleshooting guidance.
{
"mcpServers": {
"elastio": {
"command": "npx",
"args": ["-y", "https://mcp.elastio.com/elastio-mcp-0.9.0.tgz"],
"env": {
"ELASTIO_TENANT_URL": "https://your-tenant.app.elastio.com",
"ELASTIO_API_KEY": "your-access-token"
}
}
}
}For additional information and troubleshooting, see the Claude Desktop MCP documentation.
To set up the Elastio MCP server with Claude Code CLI:
# Add the Elastio MCP server
claude mcp add elastio -e ELASTIO_TENANT_URL=https://your-tenant.app.elastio.com \
-e ELASTIO_API_KEY=your-access-token \
-- npx https://mcp.elastio.com/elastio-mcp-0.9.0.tgz
# Verify the server was added
claude mcp listFor additional information and troubleshooting, see the Claude Code MCP documentation.
To set up the Elastio MCP server with Cursor:
Create a configuration file at either:
.cursor/mcp.json in your project directory~/.cursor/mcp.json in your home directoryAdd the following configuration:
{
"mcpServers": {
"elastio": {
"command": "npx",
"args": ["-y", "https://mcp.elastio.com/elastio-mcp-0.9.0.tgz"],
"env": {
"ELASTIO_TENANT_URL": "https://your-tenant.app.elastio.com",
"ELASTIO_API_KEY": "your-access-token"
}
}
}
}For additional information and troubleshooting, see the Cursor MCP documentation.
To set up the Elastio MCP server in VS Code:
Enable MCP support in settings:
chat.mcp.enabled to trueCreate a configuration file at .vscode/mcp.json with:
{
"inputs": [
{
"type": "promptString",
"id": "elastio-api-key",
"description": "Elastio API Key",
"password": true
}
],
"servers": {
"Elastio": {
"type": "stdio",
"command": "npx",
"args": ["-y", "https://mcp.elastio.com/elastio-mcp-0.9.0.tgz"],
"env": {
"ELASTIO_TENANT_URL": "https://your-tenant.app.elastio.com",
"ELASTIO_API_KEY": "${input:elastio-api-key}"
}
}
}
}For additional information and troubleshooting, see the VS Code MCP documentation.
For setup instructions with Windsurf, please refer to the Windsurf MCP documentation.
For other MCP hosts, consult their documentation for how to add custom MCP servers. The Elastio MCP server can be used with any host that supports the Model Context Protocol.
The command to run the Elastio MCP server is npx -y https://mcp.elastio.com/elastio-mcp-0.9.0.tgz. Either pass
the requirement configuration parameters as environment variables or command line arguments, using whatever mechanism
your specific MCP host supports.
Latest: v0.9.0 (Released: 2025-07-30)
The MCP server is evolving quickly; prior versions should not be used as they may not be compatible with the latest API endpoints and MCP features.
Unified threat detection experience - MCP now provides consistent threat detection capabilities across both AWS and on-premises environments, with proper entity type handling for complex assets like VMs with multiple volumes.
Enhanced API routing - Threat APIs now automatically route requests to appropriate endpoints based on entity type (asset vs asset item), ensuring correct results for all asset types.
Enhanced MCP logging integration - Use MCP logging to log errors from tool calls to the MCP host for better debugging and monitoring
Updated MCP SDK - Updated MCP SDK from version 1.13.0 to 1.13.3 for improved compatibility and performance, in particular fixing SDK bug in which error responses were expected to adhere to the tool's output schema.
Output schemas for all MCP tools - Introduced comprehensive output schemas for all tools using the latest MCP SDK specification. This enables LLMs to make better decisions about tool invocation and output processing.
Documentation search tool prototype - Added initial prototype for searching Elastio documentation, though currently disabled pending full implementation.
Updated MCP SDK to version 1.13.0 - Updated to the latest MCP SDK and refactored tool definitions to use the top-level 'title' property instead of 'annotations.title' as per SDK recommendations.
Enhanced typed tool output support - Added helper functions for tools that produce typed output, improving type safety and developer experience across all tools.
Improved MCP resource handling - Added documentation explaining why no MCP resources are defined and improved overall MCP integration patterns.
Built-in prompts for MCP server - Introduced three built-in prompts that enhance the MCP server's ability to guide users and provide contextual assistance.
Improved package execution - Removed cli.js and fixed npx execution to make the package easier to install and run across different environments.
Bundle size optimization - Added minification to the build process to reduce bundle size and improve load times while discouraging reverse engineering.
Updated MCP SDK to latest version - Upgraded to the latest TypeScript MCP SDK for improved compatibility and new features.
Enhanced error handling - Improved error handling throughout the application to provide more detailed and useful error responses.
URL prefix configuration - Added configurable URL prefix support with '/gateway' as the new default, and fixed related type checking errors.
Deployment infrastructure improvements - Added Terraform configuration for infrastructure management and improved deployment scripts.
The results produced by large language models (LLMs) in this preview should be reviewed and verified by human operators, especially before taking any critical actions as part of an incident response workflow.
While the MCP tooling itself is deterministic, meaning that the raw data retrieved from Elastio will always match what is shown in the UI, the LLM operates independently and may misinterpret or misrepresent information based on how it processes the data.
Note that Elastio is not processing any of your data using LLMs; this MCP tool will be using the metadata of the scans provided by Elastio and using LLMs on your end to answer the questions.
We welcome and encourage feedback, particularly in cases where the model provides incorrect or misleading responses. In many cases, we will be able to improve the system's reliability by refining how we structure prompts and tool interactions to mitigate LLM misbehavior over time.