mcp -rhoai-workbenches

This MCP server provides a centralized interface for managing Red Hat OpenShift AI (RHOAI) workbenches, images, hardware profiles, and storage within an OpenShift cluster. It supports listing and querying operations by default and can be extended with write capabilities when configured in read-write mode. The server exposes tools to inspect resource consumption at various scopes (per workbench, per user, per namespace, or cluster-wide) and to manage resources such as workbenches, images, hardware profiles, and PVCs. Use it to integrate natural language prompts with concrete OpenShift actions, enabling smooth governance and monitoring of AI workloads in your cluster.

Prerequisites:

• Go 1.24.0 or later installed on your machine
• Access to an OpenShift cluster with RHOAI installed
• oc CLI logged in to your cluster
• Optional: golangci-lint for code linting

Installation steps:

1. Clone the repository and enter the project directory git clone https://github.com/amaly/mcp-server-rhoai.git cd mcp-server-rhoai

2. Fetch dependencies go mod download

3. Ensure authentication with OpenShift oc login <your-cluster-url>

4. Build the project make build

5. (Optional) Run tests make test

6. Run the server (example using the built binary in read-only mode) ./mcp-server-rhoai

Notes:

• The project includes a Makefile with convenient targets for build and test workflows.
• The MCP server relies on your current OpenShift login for permission-scoped operations.

Tips and conventions:

• Start in read-only mode to explore available tools and verify permissions before enabling write operations.
• The MCP mode is controlled via the MCP_RHOAI_MODE environment variable. If not set or set to anything other than 'write', the server operates in read-only mode by default.
• When integrating with editors or CLIs, you can reference the provided commands and ensure the server binary path matches your deployment.
• Ensure oc is configured and logged in to the correct cluster context to avoid permissions issues.
• If you encounter linting or test failures, you can run golangci-lint locally (if installed) and adapt the code accordingly before re-running make build.