[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$f3qslmp1909s0f":3},{"code":4,"message":5,"data":6,"timestamp":38},10000,"success",{"full_name":7,"position":8,"description":9,"language":10,"stars":11,"forks":12,"stars_gained":13,"age_days":14,"hn_threads":15,"one_liner":22,"summary_md":23,"topic":24,"github_topics":27,"license":28,"homepage":29,"updated_at":30,"history":31},"aws\u002Fagent-toolkit-for-aws",1,"Official, AWS-supported MCP servers, skills, and plugins to help AI agents build on AWS","Python",3227,403,13,184,[16],{"id":17,"title":18,"points":19,"comments":20,"posted_at":21},48448344,"Agent Toolkit for AWS",5,0,"2026-06-08","Gives AI coding agents AWS-aware tools, skills and guardrails so they can build, deploy and manage cloud applications.","A practical choice if your coding agent already writes code but stalls at the AWS boundary—service selection, deployment, and account-aware CLI calls. It is official AWS packaging, so the interesting question is whether your agent is supported, not whether the content exists.\n\n## The problem\n\nA developer asks an AI coding agent to add a serverless endpoint, then deploy it. The agent may produce plausible CDK or CloudFormation, but it does not know which AWS services fit, how the account is configured, or what the deployment steps should be in this environment. The result is a plausible-looking change that fails at the first real API call, or a deployment the developer has to reconstruct manually.\n\nThe toolkit changes that step: instead of the agent guessing, it can search AWS documentation at runtime, call AWS APIs through the AWS MCP Server, and follow bundled skills for service selection, CDK\u002FCloudFormation, serverless, containers, storage, observability, billing and SDK usage. The developer gets an agent that can carry a task from code to deployment using environment-specific information, rather than an agent that stops at generated files.\n\n## What's different\n\nThis is not a single MCP server. It ships three plugins—`aws-core` (the starting point), `aws-agents` and `aws-data-analytics`—each bundling an AWS MCP Server configuration plus agent skills. For Kiro and fx, the README is explicit that these are two independent parts: the MCP server gives runtime API access and documentation search, while local skills give task-specific guidance. Skills do not require the MCP server, and the server does not serve locally installed skills. That matters because it lets you adopt just the guidance first, before granting an agent live API access.\n\nThe install path is unusually broad: AWS CLI, Claude Code's official Anthropic marketplace, Codex, Cursor team marketplaces, Devin, Kiro, fx, and a generic `npx skills add` route for other agents. The trade-off is that each surface has its own setup and its own update mechanism. Cursor installs from a repository marketplace file; Devin can install per-personal or local; Claude Code users may need to refresh the marketplace index before a plugin is found. None of those are difficult, but the agent you use determines what you actually configure and maintain.\n\n## Who should try it\n\nGood fit: a developer already using Claude Code, Codex, Cursor, Kiro, Devin or fx on real AWS projects, especially where the agent needs to move from writing infrastructure code to deploying and inspecting resources. It is most valuable when AWS service choice and account-specific calls are the recurring friction.\n\nPoor fit: teams whose agents work fine on local code and rarely touch AWS, or those who want one universal install across an unsettled set of agent tools. The setup is per-agent and partly per-project, so a mixed-tool team will maintain several configurations. Anyone who needs only documentation search can start with the MCP server alone.\n\n## Requirements and limitations\n\nThe generic skills route requires `uv` installed. An AWS account with locally configured credentials is needed for API calls and script execution, but not for documentation search or skill discovery. That split is useful: you can evaluate the documentation side before configuring credentials for an agent that will act on your account.\n\nOngoing care centers on the agent surfaces themselves. Each install path has its own commands and plugin indexing, and the README warns that Claude Code may need a marketplace index update. The toolkit provides tools and guardrails, not a guarantee that agent actions on your account are safe; that remains an operational concern for whoever configures credentials.\n\n## HotRepo's take\n\nHotRepo recorded the repository on the daily board on 2026-10-24 and 2026-10-25, at 3,119 and 3,159 stars (+11 and +13 that day). That is two dated observations, not evidence of momentum; persistence cannot be judged. The decision point is compatibility: check the [get-started guide for the AWS MCP Server](https:\u002F\u002Fdocs.aws.amazon.com\u002Fagent-toolkit\u002Flatest\u002Fuserguide\u002Fgetting-started-aws-mcp-server.html) against the specific agent you use before committing a team to a plugin path.",{"slug":25,"name":26},"developer-tools","Developer Tools",[],"Apache-2.0",null,"2026-10-25T03:18:48Z",[32,36],{"date":33,"stars":34,"stars_gained":35},"2026-10-24",3119,11,{"date":37,"stars":11,"stars_gained":13},"2026-10-25",1792917753]