Skip to main content

Set Up with an AI Agent

You can use an AI coding agent to generate your entire deployment setup: infrastructure, CI/CD pipeline, IAM policies, and application scaffolding. The agent uses the DevRamps MCP tools to produce a working pipeline from a single conversation.

This is the fastest way to get started, especially if you're new to AWS or don't want to write Terraform and pipeline YAML by hand.

What You Need​

  • An AWS account with admin access
  • Node.js 18+
  • An MCP-compatible AI coding agent (Claude Code, Cursor, Codex, Windsurf, or similar)
  • A GitHub repo (can be empty)

1. Initialize Agent Tooling​

From your repo root:

npx @devramps/cli init-agent

This creates four files:

FilePurpose
.mcp.jsonRegisters the DevRamps MCP server with your AI agent
DEVRAMPS_AGENTS.mdRules the agent follows (terraform structure, pipeline format, etc.)
CLAUDE.mdLoads the rules into Claude Code / Codex
AGENTS.mdLoads the rules into Cursor / Windsurf / others

Restart your AI agent after running this so it picks up the MCP server.

2. Authenticate​

npx @devramps/cli login

This opens a browser for OAuth authentication and stores your credentials in ~/.devramps/configuration.json. The agent's MCP tools use these credentials automatically.

3. Ask the Agent to Set Up Your Project​

Open your AI agent in the project directory and describe what you want:

Create a Node.js API with a React frontend, deployed to AWS

The agent will:

  1. Ask a few questions -- project name, AWS account ID, region, architecture tier
  2. Generate Terraform -- VPC, ECS, ALB, S3, CloudFront, split across separate .tf files
  3. Validate the Terraform -- runs terraform validate to catch syntax errors
  4. Show a cost estimate -- line-item breakdown with optional upgrades
  5. Generate the pipeline -- .devramps/<name>/pipeline.yaml with build, deploy, and bake steps
  6. Generate IAM policies -- .devramps/<name>/aws_additional_iam_policies.json based on your Terraform resources
  7. Validate the pipeline -- checks the YAML against the DevRamps API
  8. Generate starter code -- Dockerfile, Express server, React app, etc.

4. Review and Bootstrap​

Look through the generated files. The Terraform and pipeline definition are the most important things to verify.

When you're satisfied:

npx @devramps/cli bootstrap

This provisions IAM roles, ECR repositories, S3 buckets, and a Terraform state bucket in your AWS account.

5. Deploy​

Connect your repo in the DevRamps dashboard, then push:

git add .
git commit -m "Initial project setup"
git push origin main

The push triggers your first pipeline run. The agent can monitor the deployment for you -- just ask it to check the pipeline status.

6. Debug with the Agent​

If a deployment fails, the agent has tools to investigate:

My pipeline is failing, can you debug it?

The agent can check pipeline state, pull step logs, identify the failure, and suggest fixes. It can also retry failed stages for you.

Available MCP Tools​

The DevRamps MCP server gives the agent access to these tools:

ToolWhat it does
scaffold-pipelineGenerate a pipeline.yaml file
validate-pipelineValidate pipeline YAML against the DevRamps API
generate-iam-policiesWrite IAM policies to the pipeline directory
list-pipelinesList all pipelines with status
get-pipeline-stateGet detailed state of a specific pipeline
get-step-logsPull logs for failed or running steps
retry-stageRe-run a failed stage
cancel-stageCancel an in-progress deployment

The first three work without authentication. The rest require npx @devramps/cli login.

What's Next​