Quick start

This guide gets Infrawise connected to Claude Code and talking to your AWS environment in four steps. By the end, your AI assistant can call all 15 Infrawise MCP tools to inspect infrastructure, identify issues, and suggest fixes — without you leaving your editor.

If you haven’t installed Infrawise yet, start with the installation guide.

Step 1: Configure your AWS credentials

Infrawise uses your existing AWS credentials — the same ones the AWS CLI uses. No extra IAM setup is needed beyond read-only access to the services you want to inspect.

aws configure
# or set environment variables:
export AWS_ACCESS_KEY_ID=...
export AWS_SECRET_ACCESS_KEY=...
export AWS_DEFAULT_REGION=us-east-1

For the minimum required IAM permissions, see AWS setup.

Step 2: Create infrawise.yaml

Run infrawise init in your project directory to generate the config file interactively:

infrawise init

Or create infrawise.yaml manually:

services:
  aws:
    region: us-east-1
    # profile: my-profile   # optional — uses default credential chain if omitted
  databases:
    postgres:
      - host: localhost
        port: 5432
        database: myapp
        user: postgres
        password: ${POSTGRES_PASSWORD}

See the configuration reference for all available options.

Step 3: Start Infrawise

infrawise start --claude

Infrawise scans your infrastructure, runs 27 analyzers, and writes .mcp.json to the current directory. Open Claude Code in the same directory — all 15 MCP tools are immediately available.

Tip

Using Cursor instead? Run infrawise start --cursor to write .cursor/mcp.json and open Cursor directly.

Step 4: Ask your AI assistant

In Claude Code, try:

What infrastructure issues should I fix first?

Claude Code calls get_infra_overview and returns a prioritized list of findings across your AWS services and databases.

Step 5: Drill into a specific function

Once you have the overview, follow up with function-level analysis:

Analyze the processOrder function for infrastructure issues.

Claude Code calls analyze_function with { "function": "processOrder" } and returns the services it accesses, its trigger event shape (e.g. event.Records[0].body for SQS-triggered handlers), and any related findings. This is the most useful tool when writing or reviewing Lambda handlers.

---

FAQ

What does “analyzing infrastructure” mean?

Infrawise connects to your configured AWS services using read-only API calls, extracts resource metadata (table definitions, queue configs, Lambda settings, etc.), and runs 27 rule-based analyzers against that metadata. Analyzers check for patterns like missing DLQs, default Lambda memory, disabled secret rotation, and IaC drift. No data — database rows, log messages, secret values — is ever read.

How long does the first run take?

Typically 5 to 15 seconds for a mid-sized AWS environment. Time scales with the number of resources (Lambda functions, DynamoDB tables, SQS queues, etc.) because each resource requires at least one AWS API call. Subsequent calls from your AI editor hit the in-memory cache and return instantly.

Do I need to run infrawise start every time?

Only when your infrastructure changes. infrawise start writes a static .mcp.json that points your editor to the Infrawise stdio server — the editor manages the server process from there. If you add new AWS resources, run infrawise start again (or infrawise analyze) to refresh the graph. For a long-running server that stays in sync, use infrawise dev instead.