Installation

Infrawise is an open-source MCP (Model Context Protocol) server that gives AI coding assistants like Claude Code and Cursor a real-time, read-only map of your AWS infrastructure. It runs 27 rule-based analyzers and exposes 15 MCP tools your AI can call while you write code — without leaving your editor. This page covers installation, connecting to your editor, and the AWS credentials Infrawise needs.

Prerequisites

• Node.js 18 or later — check with node --version
• AWS credentials configured — the same credentials the AWS CLI uses; see AWS setup for the minimum IAM policy
• Claude Code or Cursor — Infrawise writes the MCP config file your editor reads automatically

Install

npm install -g infrawise

Verify the installation:

infrawise --version

Connect to Claude Code

infrawise start --claude

Infrawise scans your AWS infrastructure, runs the 27 analyzers, and writes .mcp.json to the current directory. Claude Code reads .mcp.json automatically on next launch — no plugin, no extension, no manual config editing required. All 15 MCP tools are immediately available.

Connect to Cursor

infrawise start --cursor

Writes .cursor/mcp.json and opens Cursor. All 15 Infrawise MCP tools appear in Cursor’s MCP panel.

HTTP transport mode

If you prefer HTTP over stdio, or need multiple tools to share one running server instance:

infrawise dev --config infrawise.yaml

The server starts at POST http://localhost:3000/mcp and stays running in the foreground. Useful for debugging tool calls or integrating with custom MCP clients.

AWS credentials

Infrawise reads your existing AWS credentials — the same credential chain the AWS CLI uses:

# Option 1: interactive setup stored in ~/.aws/credentials
aws configure

# Option 2: environment variables
export AWS_ACCESS_KEY_ID=...
export AWS_SECRET_ACCESS_KEY=...
export AWS_REGION=us-east-1

Infrawise needs read-only access to the AWS services you configure. See AWS setup for the exact minimum IAM policy and how to scope it to specific resources.

Note

No AWS account yet? The LocalStack demo lets you run Infrawise against emulated AWS services locally using Docker — no real AWS account or credentials required.

---

FAQ

Does Infrawise work without an AWS account?

Yes. The LocalStack demo runs Infrawise against locally emulated AWS services using Docker. You need a free LocalStack account (no credit card) but no real AWS account.

Does Infrawise store my infrastructure data?

No. Infrawise runs entirely on your local machine. Infrastructure data is held in memory during the session and is never sent to any external server. There is no telemetry and no cloud sync.

Can I use Infrawise with VS Code?

Not directly. Infrawise targets editors with native MCP support: Claude Code and Cursor. For HTTP transport use cases, infrawise dev starts a server at POST http://localhost:3000/mcp that any MCP-compatible client can connect to.

What happens when I run infrawise start?

infrawise start does three things in sequence: reads infrawise.yaml in the current directory, scans the configured AWS services and databases using read-only API calls, runs the 27 analyzers to generate findings, writes the MCP config file for your editor (.mcp.json for Claude Code, .cursor/mcp.json for Cursor), then exits. The scan typically completes in a few seconds depending on the number of resources in your account.

Why do some security scanners flag Infrawise?

Some supply-chain security scanners flag this package under “deceptive naming” because of the prefix “infra.” This is a false positive from automated tooling. Infrawise is completely safe, open-source, and unaffiliated with any commercial trademarks. You can verify by reading the source on GitHub.