You can set up an AWS S3 bucket directly inside Claude Code by prompting it to scaffold the AWS CLI commands, bucket policies, IAM roles, and infrastructure-as-code files you need. It works best when you already have the [AWS CLI installed](https://docs.aws.amazon.com/cli/latest/userguide/getting-started-install.html) and credentials configured. Claude Code handles the scaffolding, policy JSON, and SDK calls so you spend less time reading docs and more time shipping.

- Requires: AWS CLI v2, an IAM user or role with S3 permissions, and a Claude Pro or Max plan
- Average setup time with Claude Code: under 5 minutes for a standard bucket with versioning and public-access blocking
- Claude Code can generate Terraform, CDK, CloudFormation, or raw CLI commands depending on your stack

## What you need before starting

Before asking Claude Code to scaffold anything, make sure the following are in place in your terminal environment:

- **AWS CLI v2** installed and on your `$PATH` (run `aws --version` to confirm)
- **Credentials configured** via `aws configure` or an assumed IAM role. Claude Code reads from your shell environment, so whatever is active in your terminal is what it will use.
- **A Claude Pro or Max subscription**. S3 setup tasks involve multiple back-and-forth steps; you will burn through free-tier messages quickly. See [using Claude Code with Pro or Max](https://support.claude.com/en/articles/11145838-using-claude-code-with-your-pro-or-max-plan) for plan details.
- **A target AWS region** in mind (e.g., `us-east-1`, `eu-west-2`). S3 bucket names are globally unique, so pick your region before you start.

## How to set up an S3 bucket using Claude Code: step by step

### 1\. Open Claude Code in your project directory

Launch Claude Code from the terminal at the root of your project. This gives it context about your stack (Node, Python, Go, etc.) and any existing infrastructure files like `terraform/` or `cdk.json`.

### 2\. Describe what you need in plain language

Be specific with your prompt. The more context you give, the better the output. Examples that work well:

- _"Create an S3 bucket called my-app-assets in us-east-1. Block all public access. Enable versioning. Tag it with env=production."_
- _"Set up an S3 bucket for static website hosting with a bucket policy that allows public read, and output the CLI commands to do it."_
- _"Generate a Terraform module for an S3 bucket with server-side encryption using SSE-S3, lifecycle rules to move objects to Glacier after 90 days, and an IAM policy for a Lambda function to write to it."_

### 3\. Let Claude Code generate the commands or IaC files

Claude Code will produce one of the following depending on what you ask for:

- **Raw AWS CLI commands**: `aws s3api create-bucket`, `aws s3api put-bucket-versioning`, `aws s3api put-public-access-block`, etc. These are ready to copy-paste and run.
- **Terraform HCL**: An `aws_s3_bucket` resource block with associated `aws_s3_bucket_versioning`, `aws_s3_bucket_server_side_encryption_configuration`, and IAM policy documents.
- **AWS CDK (TypeScript or Python)**: A `s3.Bucket` construct with props for encryption, versioning, removal policy, and CORS if needed.
- **CloudFormation YAML**: A `AWS::S3::Bucket` resource with properties and an optional `AWS::S3::BucketPolicy`.

### 4\. Review the bucket policy and IAM output carefully

Claude Code is good at generating policies, but always review them before applying. Common things to double-check:

- The `Principal` in any bucket policy (avoid `"*"` unless you explicitly need public access)
- CORS rules, especially if you are setting up a frontend that calls S3 directly
- The `DeletionPolicy` in Terraform or CDK (defaulting to `Destroy` in dev, `Retain` in production)

### 5\. Run or apply the output

For CLI commands, execute them directly in your terminal. For Terraform, run `terraform init` and `terraform apply`. Claude Code can also write a short shell script wrapper if you want to parameterize the bucket name or environment.

### 6\. Verify the bucket

Ask Claude Code to generate a quick verification snippet: `aws s3 ls s3://your-bucket-name` or a small SDK script that uploads a test file and reads it back. This confirms permissions are correct end-to-end.

## Useful slash commands for this workflow

Claude Code's [slash commands](/content/blog/how-to-use-claude-code-slash-commands/index.html) speed up the S3 setup process:

- `/init`: Sets up a `CLAUDE.md` file in your project so Claude Code knows your AWS region, preferred IaC tool, and naming conventions from the start
- `/clear`: Resets context if the conversation has drifted and you want a clean slate for a new bucket or policy
- `/usage`: Shows your current Claude Code message usage so you do not get cut off mid-task. See the full [Claude Code slash commands list](/content/blog/claude-code-slash-commands-list/index.html) for more.

## Common S3 configurations Claude Code handles well

| Use case | What to ask for | Key resource/option |
| --- | --- | --- |
| Static website hosting | "Set up S3 for static site with CloudFront" | `aws_s3_bucket_website_configuration` |
| Private app assets | "Bucket with presigned URL access only" | Block public access + IAM role |
| Data lake / log storage | "Bucket with lifecycle rules and Glacier tiering" | `aws_s3_bucket_lifecycle_configuration` |
| Cross-account access | "Bucket policy to allow reads from account 123456" | Resource-based policy with `Principal.AWS` |
| Replication | "Replicate this bucket to eu-west-1 for DR" | `aws_s3_bucket_replication_configuration` |

## Avoiding the mid-task usage lockout

S3 setup in Claude Code often involves several rounds: create the bucket, fix the policy, add encryption, wire up IAM. That's enough back-and-forth to push you close to your hourly usage limit. [Claude's usage limits](https://support.claude.com/en/articles/9797557-usage-limit-best-practices) reset on a rolling 5-hour window, and hitting the cap mid-task means waiting with an incomplete setup.

You can also type `/usage` inside Claude Code at any point to get a quick read on where you stand.

## Tips for better results

- **Provide your IaC tool upfront**: Say "using Terraform 1.7" or "using AWS CDK v2 TypeScript" so Claude Code doesn't guess. This avoids deprecated resource names and mismatched API versions.
- **Mention existing infrastructure**: If you already have a VPC, an existing IAM role, or a naming convention like `company-env-service`, tell Claude Code. It will reference the right resources instead of creating duplicates.
- **Ask for outputs**: "Also output the bucket ARN and name as Terraform outputs" saves you from writing boilerplate yourself.
- **Iterate in one session**: Keep the conversation in one Claude Code session. It maintains context across follow-up prompts, so you can say "now add a CORS rule for [https://myapp.com](https://myapp.com/)" and it will edit the existing file rather than starting over.
- **Use `/init` to document conventions**: If you run multiple S3 setup tasks across projects, the `CLAUDE.md` file generated by `/init` lets you [define persistent context](/content/blog/how-to-setup-claude-code-hooks-tutorial/index.html) so Claude Code always knows your preferred patterns.

## Key takeaways

1. Install and configure the AWS CLI before starting; Claude Code inherits your shell credentials.
2. Specify your IaC tool (Terraform, CDK, CloudFormation, or raw CLI) in your first prompt.
3. Use descriptive prompts: bucket name, region, access pattern, encryption, and tagging in one message.
4. Always review generated IAM policies and bucket policies before applying them.
5. Use `/usage` or [Usagebar](/content/site-root.html) to track your message budget so a multi-step S3 task doesn't get cut off mid-run.
6. Use `/init` to store AWS conventions in `CLAUDE.md` for repeatable, consistent results.