Skip to content

02 · Secure AWS landing zone

Outcome: a multi-AZ AWS VPC with public, private and isolated data subnets, a NAT gateway, encrypted flow logs, a hardened bastion reachable only from your IP, an account security baseline (CloudTrail, GuardDuty, Access Analyzer) and remote Terraform state in a hardened S3 bucket. Then you change it, update your IP, remove one part and tear it all down.

Verified with --dry-run (Terraform render + validate); run for real with cloud credentials

tests/scenarios/02-aws-landing-zone.sh runs every command below that works without an AWS account (render, terraform validate, estimate, status, targeted and full destroy of the rendered environment) and checks that the credentialed ones stop with a clear message. --plan-only, the apply, ssh and update-ip need your AWS credentials.

Time Cost Level Needs
~25 min (apply ~10 min) ≈ $58 / month on-demand list prices (cs finops estimate), mostly the NAT gateway Intermediate An AWS account

What you'll build

%%{init: {"flowchart": {"nodeSpacing": 30, "rankSpacing": 38, "padding": 12}}}%%
flowchart TB
  you["Your IP only<br/>(--allow-ip)"] -- "SSH :22" --> bastion
  subgraph vpc["VPC 10.0.0.0/16 (2 availability zones)"]
    subgraph pub["Public subnets"]
      bastion["Bastion<br/>AL2023, IMDSv2, KMS disk, SSM"]
      nat["NAT gateway"]
    end
    subgraph priv["Private subnets"]
      wl["Your workloads<br/>workload_security_group_id"]
    end
    subgraph data["Isolated data subnets"]
      db["Databases<br/>(no internet route)"]
    end
  end
  wl --> nat
  bastion -- "SSH" --> wl
  subgraph baseline["Security baseline"]
    ct["CloudTrail (multi-region, KMS)"]
    gd["GuardDuty + Access Analyzer"]
    fl["VPC flow logs to CloudWatch (KMS)"]
  end
  state["Remote state<br/>S3: versioned, SSE-KMS, TLS-only, lockfile"]

Before you start

  • An AWS account and a way to log in: aws configure, aws sso login, or a named profile.
  • Terraform >= 1.10. Step 1 shows three ways to get it.
  • The account baseline (CloudTrail, S3 public-access block, IAM password policy) is a singleton: if another environment in this account already owns it, add --var enable_account_baseline=false (AWS reference).

Step 1: Get the tools (three ways)

cs deps install terraform aws
cs deps status

Homebrew when present; otherwise Terraform's official release (SHA256-verified) into ~/.cloudseed/bin and the AWS CLI from AWS's installer. cs deps install all adds gcloud and az.

cs deps image
cs --runtime container setup aws -y --env prod --region us-west-2 --allow-ip 203.0.113.7 --dry-run
cs deps runtime container

Builds cloudseed:local (Terraform, aws, gcloud, az, kubectl, helm) with Docker or Podman. --runtime container runs one command in it; deps runtime container makes it the default. CLOUDSEED_HOME, ~/.aws, ~/.config/gcloud and ~/.azure are mounted. Back to local tools: cs deps runtime auto.

cs deps bundle

Builds dist/cloudseed-<os>-<arch>: the CLI, every Terraform module and a verified Terraform release in one file, for machines with nothing installed.

Step 2: Log in and check

aws sso login --profile prod
cs creds set AWS_PROFILE=prod
cs doctor aws

cs creds is cloudseed's local vault (~/.cloudseed/credentials.json, 0600): its values are injected into every cloudseed command, the web console and the MCP server, and never shown to AI agents. A variable exported in your shell always wins.

Expected output (abbreviated)
  ✔ AWS_PROFILE stored in ~/.cloudseed/credentials.json

  ━━ Amazon Web Services ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━
    ✔ terraform              1.16.1                 /opt/homebrew/bin/terraform
    ✔ ssh-keygen             ok                     /usr/bin/ssh-keygen
    ✔ aws                    aws-cli/2.36.40        /opt/homebrew/bin/aws

Step 3: See every knob

cs help variables aws
cs help outputs aws
cs explain state

Every variable of the stack can be set with --var name=value (az_count, single_nat_gateway, bastion_instance_type, enable_security_hub, enable_kubernetes, enable_vpn, ...). The list is generated from terraform/aws/variables.tf, so it is always current.

Step 4: Preview offline with a dry run

Replace 203.0.113.7 with your public IP (or leave --allow-ip out: cloudseed detects it).

cs setup aws -y --env prod --region us-west-2 --allow-ip 203.0.113.7 --dry-run
Expected output (abbreviated)
  ╭─ Environment aws-prod ─────────────────────────────────────────────────────╮
  │ Cloud                     Amazon Web Services                               │
  │ Region                    us-west-2                                         │
  │ Network CIDR              10.0.0.0/16                                       │
  │ SSH allowed from          203.0.113.7/32                                    │
  │ State                     remote  (storage created on the first apply)      │
  │ enable_account_baseline   true                                              │
  │ az_count                  2                                                 │
  │ single_nat_gateway        true                                              │
  │ bastion_instance_type     t3.micro                                          │
  │ Tags                      Project=cloudseed, Environment=prod, Owner=you,   │
  │                           ManagedBy=cloudseed, CloudseedEnv=aws-prod, ...   │
  ╰─────────────────────────────────────────────────────────────────────────────╯
$ terraform validate -no-color
Success! The configuration is valid.

$ terraform validate -no-color
Success! The configuration is valid.

  ✔ Dry run complete. Rendered root(s): ~/.cloudseed/envs/aws-prod/stack, ~/.cloudseed/envs/aws-prod/bootstrap

Two roots are validated: bootstrap (the S3 state bucket, created first on apply) and stack (everything else).

Step 5: Know the bill before you apply

cs finops estimate aws --env prod
Expected output
  ╭─ Estimate · aws-prod  (month, 730h at on-demand list prices) ─────────────╮
  │ bastion t3.micro x1           $7.59                                        │
  │ NAT gateway x1                $32.85                                       │
  │ public IPv4 x2                $7.30                                        │
  │ KMS key x1                    $1.00                                        │
  │ CloudTrail + logs (low vol.)  $3.00                                        │
  │ GuardDuty (low volume)        $5.00                                        │
  │ block storage ~10 GB          $0.80                                        │
  │                                                                            │
  │ total / month                 $57.54                                       │
  ╰────────────────────────────────────────────────────────────────────────────╯

Step 6: Plan against your account, then apply

cs setup aws -y --env prod --region us-west-2 --allow-ip 203.0.113.7 --plan-only

A first --plan-only with remote state plans the state storage and the stack and creates nothing. When the plan looks right, apply:

cs setup aws --env prod --region us-west-2 --allow-ip 203.0.113.7
cs setup aws -y --env prod --region us-west-2 --allow-ip 203.0.113.7 --auto-approve

The S3 state bucket (versioned, SSE-KMS, TLS-only, native lockfile) is created first, then the stack, then the bastion is provisioned: the repository is copied over SSH and Ansible hardens it.

Step 7: Look around

cs status aws --env prod
cs output aws --env prod --json | jq -r '.private_subnet_ids[]'
cs ssh aws --env prod

Put workloads in the private subnets and attach workload_security_group_id: it accepts SSH only from the bastion.

Step 8: Your IP changed? One command

cs update-ip aws --env prod
cs update-ip aws --env prod --allow-ip 198.51.100.4,203.0.113.0/24 --auto-approve

Only the SSH-source change is applied; other pending changes stay untouched. Afterwards cloudseed checks that the bastion answers again.

Step 9: Change it, remove a part, bring it back

cs setup aws --env prod --var single_nat_gateway=false
cs plan aws --env prod
cs destroy aws --env prod --select
cs destroy aws --env prod --target module.stack.module.bastion
cs apply aws --env prod
  • Re-running setup with a new --var updates the environment in place (here: one NAT gateway per AZ), always showing the plan first.
  • destroy --select lists modules and resources as a numbered list; --target takes Terraform addresses. A partial destroy keeps the configuration, so cs apply re-creates what was removed.

Use an agent, MCP or the UI

Follow the same numbered steps and verification/cleanup conditions through your chosen interface. Start with the interface setup and coverage guide; replace account/project/subscription and SSH placeholders before any live request.

Agent prompt: “Follow the AWS landing-zone walkthrough for prod in my approved region and account. Preview setup, estimate cost and show baseline ownership. After approval deploy, verify private routing and inspect the plan before any targeted deletion.”

MCP starter: cloudseed_setup with:

{
  "cloud": "aws",
  "env": "prod",
  "region": "us-west-2",
  "allow_ip": "203.0.113.7/32",
  "dry_run": true
}

Use the matching tool for each remaining step in this page; the command-to-tool map lists the tool family. Keep aws-prod selected. Preview first; add confirm:true only to the specific change you have authorized. Host bootstrap, provider login and interactive applications retain their documented human steps.

UI: Create → AWS: enter prod, your region and real SSH source; choose Dry run first. Use Environments → prod for plan, status, IP update and destroy. Use All actions → FinOps for the estimate. Provider login and runtime installation are host bootstrap steps.

Verify it worked

cs inventory aws --env prod
cs troubleshoot aws --env prod
  • inventory lists every managed resource (VPC, subnets, NAT gateway, bastion, KMS key, CloudTrail, GuardDuty, ...) with its IDs, plus the history of applies and provisioning runs.
  • troubleshoot checks credentials, that your current IP is still allowed and that the bastion answers on port 22.
  • In the AWS console: CloudTrail shows the multi-region trail, GuardDuty is enabled, and every resource carries ManagedBy=cloudseed and CloudseedEnv=aws-prod tags.

Clean up

cs destroy aws --env prod --purge-state --purge

A full destroy shows the plan and asks you to type aws-prod. --purge-state also deletes the S3 state bucket, --purge the working directory (the audit trail is kept in ~/.cloudseed/logs/purged/aws-prod/). The account-wide S3 public-access block, IAM password policy and EBS default encryption stay in place: they protect the whole account. Unattended: add -y --auto-approve.

What just happened

cs explain reconcile
cs explain security-baseline

Next steps