← Explore all mcp servers

CLOUD OPERATIONS / AWS Labs

SageMaker AI MCP

AWS Labs MCP server for SageMaker HyperPod clusters on EKS or Slurm, intended for local stdio and read-only until write and sensitive-data flags are added.

“Manages HyperPod deployment, nodes, and lifecycle, and says not to modify the server for network use.”

01 / THE REASONING

Why this made the selection.

  • The catalogue has no SageMaker or HyperPod MCP server.
  • The README separates read-only tools from the write tools that need --allow-write.

02 / THE REVIEW RECORD

What we actually inspected.

Source review has boundaries.
A clear record is more useful than a “safe” badge.

Material inspected

  • src/sagemaker-ai-mcp-server/README.md
  • src/sagemaker-ai-mcp-server/LICENSE

Our findings

  • The current surface is SageMaker HyperPod on Amazon EKS or Slurm, including deployment, node management, and lifecycle operations.
  • The README says the server is read-only by default and names manage_hyperpod_stacks deploy and delete, plus node update_software and batch_delete, as write operations that need --allow-write.
  • It says the server is for STDIO mode only, runs with the user's permissions and complete filesystem access, and should not be modified for network use. Its directory LICENSE is Apache-2.0.

Not established by this review

  • The server was not started.
  • No HyperPod tool was called.
  • No cluster was changed.

The review applies to the material and revision named here. A newer upstream release can change its behavior.

03 / PUT IT TO WORK

Connect a server deliberately.

Upstream setup instructions ↗
  1. Client examples use uvx awslabs.sagemaker-ai-mcp-server@latest inside JSON, not a single shell line.
  2. Drop --allow-write to stay in the read-only mode the README recommends.
  3. The directory LICENSE was fetched separately and is Apache-2.0.

Before you start

  • uv
  • AWS credentials with access to the HyperPod cluster
  • A local stdio MCP client

04 / FOLLOW THE EVIDENCE

The source trail.

Our notes are separate from the original resource.
Check upstream before adopting a new version.

  1. SageMaker AI README ↗Checked

    https://github.com/awslabs/mcp/blob/8840003820a78dd018561d55c7ab12fcdd708464/src/sagemaker-ai-mcp-server/README.md

    Supports: STDIO warning and write flag

  2. Directory LICENSE ↗Checked

    https://github.com/awslabs/mcp/blob/8840003820a78dd018561d55c7ab12fcdd708464/src/sagemaker-ai-mcp-server/LICENSE

    Supports: Apache-2.0

Evidence & Ask