ZenML

ZenML

Interact with your MLOps and LLMOps pipelines through your ZenML MCP server

37
Stars
10
Forks
6
Releases

Overview

An MCP server that exposes ZenML's core read capabilities over the Model Context Protocol to enable LLMs and other tools to query live ZenML state. The server provides read access to live information about users, stacks, pipelines, pipeline runs, pipeline steps, services, stack components, flavors, pipeline run templates, schedules, and artifacts (metadata about data artifacts, not the data itself), as well as service connectors, step code, and step logs (including those from cloud-based stacks). It also supports triggering new pipeline runs when a corresponding run template is available. The project ships as MCP Bundles (.mcpb)—an all-in-one package that includes runtime dependencies and secure configuration—and as a Docker image for stdio-based communication. It is designed to work with MCP clients such as Claude Desktop and Cursor, with setup instructions for both. The repository includes automated tests and CI for reliability, and is published to the Anthropic MCP Registry so the ZenML MCP Server can be discovered and installed by compatible hosts. This release is described as beta/experimental as the integration evolves and feedback is welcomed.

Details

Owner
zenml-io
Language
Python
License
MIT License
Updated
2025-12-07

Features

Live read access to ZenML data via MCP

Provides MCP-based access to live information about users, stacks, pipelines, pipeline runs, pipeline steps, services, stack components, flavors, run templates, schedules, artifacts, service connectors, step code, and step logs.

Trigger new pipeline runs using run templates

Allows triggering pipeline runs when a run template is present.

MCP Bundles (.mcpb) for easy deployment

Ships as MCP Bundles packaging the server and dependencies into a single file, replacing older formats and simplifying distribution.

Dockerized deployment with stdio-based MCP communication

Run the server in a Docker container with stdio-based communication and environment-variable configuration for ZenML credentials.

Claude Desktop and Cursor compatibility

Includes setup guidance for Claude Desktop and Cursor MCP clients, including per-repo configuration for Cursor.

Prebuilt Docker images and local build options

Offers prebuilt multi-arch Docker images and a local build path to run the MCP server.

Automated testing and CI support

Supports automated smoke tests, fast CI, and issue creation on test failures to ensure reliability.

Audience

MCP hostsPrograms like Claude Desktop or IDEs that access ZenML data via MCP.
MCP clientsClients maintaining 1:1 MCP connections to servers for real-time data access.
ZenML developersDevelopers and data engineers integrating ZenML capabilities with LLM-based workflows.

Tags

MCPZenMLMLOpsLLMOpsPipelinesArtifactsRun TemplatesSchedulesService ConnectorsStep CodeStep LogsDockerMCP BundlesAnthropic RegistryClaude DesktopCursor