AI agent plugins, tools, and resources for startup builders on AWS.
| Plugin | Description | Status |
|---|---|---|
| migration-to-aws | Assess, plan & execute: migrate GCP/Heroku infrastructure and AI workloads to AWS (discovery, architecture mapping, cost analysis, Terraform), rewrite LLM SDK calls to Amazon Bedrock, and select an AWS runtime + build a POC for AI agents. Bundles the gcp-to-aws, heroku-to-aws, llm-to-bedrock, and agent-advisor skills | Available |
| aws-dev-toolkit | AWS development toolkit — 35 skills, 11 agents, and 3 MCP servers for building, migrating, and architecture reviews on AWS | Available |
# Add the marketplace
/plugin marketplace add awslabs/startups
# Install the plugin
/plugin install migration-to-aws@startups-for-awscodex plugin marketplace add awslabs/startups
codex plugin install migration-to-awsComing soon — Plugins are not yet published on the Cursor Marketplace.
migration-to-aws is a single plugin bundling four skills that cover assessment, execution, and agent runtime decisions:
- gcp-to-aws / heroku-to-aws — assess and plan a platform migration: scan infrastructure (Terraform, billing, source code), map services to AWS equivalents, estimate costs, and generate validated Terraform and migration scripts.
- llm-to-bedrock — execute an AI/LLM migration: rewrite your SDK calls to Amazon Bedrock's Converse API, run quality evaluation against a golden dataset, and deliver the changes on a ready-to-merge git branch. (On platforms without subagent dispatch it runs inline — slower, but fully functional.) The migration skills delegate the AI-execution step here.
- agent-advisor — decide how and where to run AI agents on AWS: deterministic runtime scoring (AgentCore / ECS / EKS / Lambda / Batch / MicroVMs), multi-workload decomposition into units, Temporal worker handling, and a layered recommendation → migration plan → deployable POC.
migration-to-aws (one plugin, four skills)
gcp-to-aws ─┐
├─▶ assess & plan ──▶ llm-to-bedrock ──▶ rewrite · evaluate · branch
heroku-to-aws ┘ (Terraform, (AI/LLM execution)
cost, scripts)
agent-advisor ──▶ score runtime ──▶ recommendation · plan · POC
(how/where to run agents on AWS)
Each top-level folder is owned by a team and contains their plugins, tools, or resources:
awslabs/startups/
├── .claude-plugin/marketplace.json # Plugin marketplace (lists all plugins)
├── migrate/ # Migration tools and plugins
│ └── plugins/
│ └── migration-to-aws/ # Assess, plan, execute + agent runtime advisor
│ # skills: gcp-to-aws, heroku-to-aws,
│ # llm-to-bedrock, agent-advisor
├── solution-architecture/ # Solution Architecture plugins (aws-dev-toolkit)
└── ... # Future team folders
To add a new plugin to the marketplace:
- Create your plugin under your team's folder (e.g.,
migrate/plugins/my-plugin/) - Include a
.claude-plugin/plugin.jsonmanifest in your plugin directory - Add an entry to the root
.claude-plugin/marketplace.json:
{
"name": "my-plugin",
"source": "./my-team-folder/plugins/my-plugin",
"version": "1.0.0",
"description": "What your plugin does"
}- Submit a PR — requires approval from
@awslabs/startups-admins(for marketplace changes) and your team's CODEOWNERS (for plugin content)
See CONTRIBUTING.md for contribution guidelines, the first-time publishing process, and documentation requirements.
See CONTRIBUTING for security issue notifications.
This project is licensed under the Apache-2.0 License. See LICENSE for details.