MCP Server for ZenML

What is MCP Server for ZenML?

MCP Server for ZenML is an MCP server that provides standardized tools for Large Language Models to interact with the ZenML API. It allows AI applications to read core entities (users, stacks, pipelines, models) and trigger pipeline runs, making live ML platform data accessible through the Model Context Protocol.

How to use MCP Server for ZenML?

The easiest way is through the ZenML dashboard's MCP Settings page, which provides pre-configured snippets and one-click installation. For manual setup, you can run the server via Docker (especially for MCP Apps with Streamable HTTP transport) or using uv. Configuration requires a ZenML server URL, API key, and optionally an active project ID.

Key features of MCP Server for ZenML

• List and get details on users, stacks, components, and flavors
• Browse pipeline definitions, runs, steps, schedules, and artifacts
• Read metadata about models, model versions, services, and deployments
• Trigger new pipeline runs using snapshots or run templates
• Diagnose server setup with environment and connectivity checks
• Open interactive dashboards via experimental MCP Apps

Use cases of MCP Server for ZenML

• Discover project context and find runnable snapshots for a pipeline
• Trigger a new pipeline run and monitor its deployment status
• Analyze recent pipeline runs and stack component usage
• Retrieve step logs, source code, and artifact metadata for debugging
• Diagnose ZenML server connectivity and authentication issues

FAQ from MCP Server for ZenML

What is MCP?

MCP (Model Context Protocol) is an open standard that lets AI applications connect to data sources and tools in a uniform way, like a “USB-C port for AI.”

What is ZenML?

ZenML is an open-source platform for building and managing ML and AI pipelines, providing a unified interface for data, models, and experiments.

What transport does the server use for MCP Apps?

MCP Apps require Streamable HTTP transport; they cannot work via stdio.

Which AI clients currently support MCP Apps?

VS Code (Insiders), Goose, and ChatGPT (launching soon). Claude Desktop and Claude.ai do not yet support Apps as of late January 2026.

How do I migrate from Run Templates to Snapshots?

Replace list_run_templates() with list_snapshots(runnable=True, named_only=True), get_run_template(name) with get_snapshot(name, include_config_schema=True), and trigger_pipeline(template_id=...) with trigger_pipeline(snapshot_name_or_id=...).