Cai Workbench MCP Server
Cloudera AI Workbench API interactions for LLMs via FastMCP
Data last scanned today · Reviewed today
Overview
Cai Workbench MCP Server bridges the gap between AI assistants and Cloudera AI Workbench instances. By leveraging the official cmlapi Python SDK, it enables LLMs to perform full lifecycle management of ML models, monitor jobs, and handle project discovery operations.
Our verdict
This server carries a high-risk rating of 35/100, primarily because the authentication methods and read-only mode support remain unreviewed. Given that it is a community-maintained project rather than an official Cloudera release, users should exercise caution. It is suitable for those familiar with CML API security, but not recommended for environments where strict, verified supply-chain security is a requirement.
- Uses official cmlapi Python SDK for reliable API calls
- Zero direct dependencies reduces external supply-chain surface area
- Preserves directory structures during file upload operations
- Supports STDIO communication for secure local subprocess integration
- HTTP transport mode lacks any authentication mechanism
- Community-maintained project lacking official vendor support
- Security review for tool-specific read-only modes is missing
Tools
| Tool | Description | Risk |
|---|---|---|
| list_projects_tool | Lists projects. | low |
| get_project_id_tool | Gets a project ID. | low |
| update_project_tool | Updates a project. | medium |
| create_project_tool | Creates a project. | medium |
| get_project_tool | Gets project details. | low |
| delete_project_tool | Deletes a project. | high |
| list_project_names_tool | Lists project names. | low |
| list_teams_tool | Lists teams. | low |
| list_project_collaborators_tool | Lists project collaborators. | low |
| add_project_collaborator_tool | Adds a project collaborator. | medium |
Compatibility
| Client | Local | Docker | Remote | Read-only |
|---|---|---|---|---|
| Claude Desktop | ||||
| Cursor | ||||
| VS Code | ||||
| Windsurf | ||||
| ChatGPT |
Frequently asked questions
›What is the recommended way to connect this MCP server to a client?
STDIO is the recommended transport mode for local applications like Claude Desktop, Agent Studio, and Cursor.
›What prerequisites are needed to run this server?
You need Python 3.10 or higher, the uv package manager, a Cloudera AI instance, and your specific API key.
›Why is the '--with' argument required when using uvx?
The --with argument is required to install the cmlapi SDK from your Cloudera AI instance; without it, the tools will fail because the required module cannot be found.
›Can I use HTTP mode for production deployments?
No, HTTP mode is intended only for development and testing because it runs without any authentication.
›How does the server communicate with Cloudera AI APIs?
The server uses the official CMLServiceApi Python SDK for all tool functions to ensure correct request formatting and type safety.
Changelog
- New
- NewRelease v2.1.0
2026-05-27
- BreakingCAI Workbench MCP Release version v2.0.0
2026-05-12
- SecurityCAI Workbench MCP Release version v1.0.0
2025-10-23
Badge
Maintain this server? Add the live badge to your README.