AWS MCP Servers
AWS's official MCP servers: sixty-plus specialised ones to pick from, plus two that AWS hosts itself
What is AWS MCP Servers?
Point Claude, Cursor, Kiro or any other MCP client at one of these servers and the model can search current AWS documentation, write CloudFormation and CDK against AWS's own guidance, query DynamoDB or PostgreSQL, read CloudWatch logs and estimate costs — each capability is its own small server, started with one uvx command and your AWS credentials. Two need no setup at all: the hosted Knowledge server answers documentation questions without an AWS account, and the preview AWS MCP Server reaches the AWS APIs under IAM permissions with CloudTrail logging. The open question is longevity: AWS announced the Agent Toolkit for AWS in May 2026 as this suite's successor, recommends it for production agent work, and says the most useful servers here will move into it over time.
What can you do with AWS MCP Servers?
- Start with the two servers AWS runs for you — The Knowledge server is a URL you paste into any MCP client: it answers from current AWS documentation, API references, What's New posts and Well-Architected guidance, and needs no AWS account at all. The AWS MCP Server — in preview — goes further and calls the AWS APIs themselves under your IAM permissions; AWS states the calls are syntactically validated and every action lands in CloudTrail.
- Pick one small server per job, not one giant one — The sixty-plus local servers each cover one service or workflow — DynamoDB, PostgreSQL, EKS, ECS, CloudWatch, pricing, Bedrock and more — and each starts with a single uvx command plus your AWS profile in the client's configuration. Enabling only the two or three you need keeps the model's tool list short enough for it to choose well.
- Infrastructure code checked against AWS's own guidance — The IaC server brings CloudFormation documentation, CDK best practices, construct examples, security validation and deployment troubleshooting into the conversation, so the model writes templates against current guidance instead of its training-data memory of AWS.
- Touching live infrastructure is a per-server decision — The operational servers separate reading from acting: explicit switches such as --allow-write, with a second one for sensitive data, decide whether a server may only inspect your account or also change it. An agent that reports on production can be wired up without the ability to modify it.
Before you choose AWS MCP Servers
- The servers you run yourself speak stdio only — AWS says streamable HTTP is still in the works — so serving a whole team from one shared endpoint is possible only with the two AWS-hosted servers.
- Individual servers get deprecated as the suite reorganises — the Cloud Control API server already points users to the IaC server instead — so a set chosen today needs re-checking on upgrades.
Star history
17 Aug to 30 Oct · +240
Frequently asked questions
Is AWS MCP Servers free for commercial use?
AWS MCP Servers is released under the Apache-2.0 licence — OSI-approved open source, which permits commercial use.
How can AWS MCP Servers be deployed?
AWS MCP Servers is available as Runs locally / Managed cloud / Self-hosted.