Terraform Infrastructure

Terraform support covers the base AgentCore agent, gateway, authentication, and frontend infrastructure. It does not project managed Intelligence credentials into the CopilotKit Runtime Lambda. Use the CDK deployment path documented in ../README.md when the managed Threads and Intelligence path is required.

Usage

Starting from the example root (examples/integrations/agentcore/):

bash

cd infra-terraform
cp terraform.tfvars.example terraform.tfvars
# Edit terraform.tfvars — set stack_name_base, backend_pattern, aws_region

terraform init
terraform plan
terraform apply

In docker mode (the default backend_deployment_type) this single apply also builds the agent's ARM64 image and pushes it to ECR before creating the runtime, so no separate build step is needed. scripts/build-and-push-image.sh exists for CI or for rebuilding the image on its own.

Testing the deployed agent

Stay in infra-terraform/ and run this directory's tester. It imports boto3, requests and colorama, which come from the example-root pyproject.toml, so run it under uv. No --project flag: uv finds the example-root project by walking up from infra-terraform/, and uv run resolves the script path against your shell's directory, not the project's — so scripts/test-agent.py is this directory's tester either way. The first run creates and syncs the example-root .venv:

bash

uv run scripts/test-agent.py 'Hello'

Deploying the frontend

scripts/deploy-frontend.py is the Terraform-side frontend deployer — it reads the Amplify app id and staging bucket from terraform output, where the example-root scripts/deploy-frontend.py is the CDK variant that reads aws cloudformation describe-stacks and takes a stack-name argument.

This path does not currently work. The script also requires a Terraform output named feedback_api_url, and no root or module outputs.tf declares one — modules/backend/ssm.tf has an SSM parameter of that name, which is not an output. So every run stops here and exits 1:

ℹ Generating aws-exports.json...
✗ Missing required Terraform outputs: feedback_api_url

It gets that far and no further: no build, no upload, no Amplify deployment. Use infra-cdk/ if you need a deployed frontend today. Fixing this — declaring the output or dropping the requirement — is tracked separately.

Once it is fixed, the invocation is:

bash

uv run --no-project scripts/deploy-frontend.py

--no-project is there because this script imports only the Python standard library, so uv has no reason to sync the example-root virtualenv; it still provisions the interpreter. uv is optional here — a plain python3 scripts/deploy-frontend.py behaves identically, needing only Python 3.8+ plus the aws, npm, node and terraform CLIs.

The agent pattern defaults to backend_pattern from terraform.tfvars. Override it with --pattern:

bash

uv run --no-project scripts/deploy-frontend.py --pattern langgraph-single-agent