Why Awnexa
Claude Code can already run the AWS CLI. So why add Awnexa? Because exploring a whole AWS system one CLI call at a time is slow, fills the conversation with raw JSON, and makes it easy for the AI to state things it never actually checked.
Without Awnexa
Section titled “Without Awnexa”When you ask Claude Code “what breaks if I delete this subnet?” with only the AWS CLI available, it has to:
- Call many APIs, one after another. Instances, network interfaces, load balancers, databases, Lambda functions, endpoints… each in every region that matters, and each with its own pagination.
- Read large JSON responses. Most of the output is irrelevant to the question, but it still takes up the conversation’s context, leaving less room for the actual reasoning.
- Guess what it did not look at. A failed or denied call shows its error once, in a tool output, and the final answer rarely carries it; a call that was never made leaves no trace at all. “I found no databases in that subnet” can quietly mean “I did not check RDS”.
- Use your credentials directly. The AI runs commands with whatever access your profile has. Read-only access depends on how carefully that profile was set up.
With Awnexa
Section titled “With Awnexa”Awnexa scans the account once with read-only APIs and saves the result on your machine. Claude Code then asks Awnexa instead of AWS:
| What Claude Code needs | What Awnexa gives it |
|---|---|
| The big picture | awnexa overview: entry points, what sits behind them and how resources group into stacks, in one call |
| Dependencies and impact | awnexa impact: what depends on a resource in the scan, with the path and how each link was established |
| What was not checked | awnexa coverage: every API and region with a status such as complete, partial, access_denied, failed or unsupported |
| What changed | awnexa snapshot diff: created, modified and deleted resources and relationships between two scans (deletions from partial scans stay unknown) |
| Just the details that matter | --fields and --limit return only the properties and rows asked for |
| Where a fact came from | Every resource and relationship keeps its source API and a confidence level such as direct, derived or inferred |
What you get:
- Fewer, smaller calls. Questions about the whole system are answered from a local database instead of many API calls, so more of the conversation is left for thinking.
- Answers that separate fact from guess. Claude Code can say “Awnexa found these three resources (scan of 10:42)” and “this part was not checked: RDS was access denied” — and keep its own reasoning apart from both.
- Read-only by design. Awnexa only calls read APIs (
Describe*,List*,Get*,BatchGet*,Search*) and never reads secret values. After the scan, Claude Code answers from the local results and needs no AWS access; whether it may still run the AWS CLI is up to your Claude Code permissions.
What Awnexa does not do
Section titled “What Awnexa does not do”Awnexa collects, organizes and compares configuration. It does not review your setup, suggest
changes, write IaC or find the cause of an incident — Claude Code does that, using the facts
Awnexa provides. Awnexa does not read logs or data, and reads metrics only as an optional usage
summary (scan --metrics). See How it works.
- Get started: install and run your first scan.
- Use cases: example conversations for six common tasks.