Tools Reference¤

The CMF MCP Server exposes a comprehensive set of tools that enable AI assistants to interact with CMF Server metadata. All tools follow a consistent pattern and return structured JSON responses.

Tool Categories¤

Tools are organized into four functional categories:

Pipeline Tools - List and explore pipelines
Execution Tools - Retrieve execution details and lineage
Artifact Tools - Discover and analyze artifacts
Additional Tools - Model cards and specialized queries

Response Format¤

All tools return responses as a list of dictionaries, allowing for multi-server queries:

[
  {
    "cmfClient": "http://cmf-server-url:8080",
    "data": [ /* CMF API results */ ]
  }
]

When querying multiple CMF Servers, results from each server appear as separate entries in the list.

Pipeline Tools¤

cmf_show_pipelines¤

Lists all pipelines available in the configured CMF Server(s).

Description: Retrieves a comprehensive list of all pipelines in your CMF infrastructure, providing an overview of available ML workflows.

Parameters:

Parameter	Type	Required	Description
`cmfClient_instances`	List[str]	No	Specific CMF Server URLs to query. If omitted, queries all configured servers.

Returns: List of pipelines with their metadata including pipeline names and properties.

Example Usage:

Show me all pipelines in CMF

List pipelines from the production CMF server

Sample Response:

[
  {
    "cmfClient": "http://server:8080",
    "data": [
      {"name": "Test-env", "id": "1", ...},
      {"name": "training-pipeline", "id": "2", ...}
    ]
  }
]

See Also: cmf_show_executions to explore executions within a pipeline

Execution Tools¤

cmf_show_executions¤

Retrieves detailed execution information for a specific pipeline.

Description: Provides comprehensive details about all executions within a pipeline, including execution parameters, timestamps, and metadata.

Parameters:

Parameter	Type	Required	Description
`pipeline`	string	Yes	Name of the pipeline to query
`cmfClient_instances`	List[str]	No	Specific CMF Server URLs to query

Returns: Detailed execution data including execution UUIDs, names, properties, and relationships.

Example Usage:

Show me all executions for the Test-env pipeline

Get execution details for the training-pipeline

Sample Response:

[
  {
    "cmfClient": "http://server:8080",
    "data": [
      {
        "Execution_uuid": "abc123...",
        "Execution": "train_model",
        "Context_Type": "train",
        ...
      }
    ]
  }
]

See Also: cmf_execution_lineage for lineage analysis

cmf_show_executions_list¤

Lists execution names for a pipeline (brief list format).

Description: Returns a concise list of execution names without full details, useful for quick discovery or when you need just the execution identifiers.

Parameters:

Parameter	Type	Required	Description
`pipeline`	string	Yes	Name of the pipeline to query
`cmfClient_instances`	List[str]	No	Specific CMF Server URLs to query

Returns: Brief list of execution names and UUIDs.

Example Usage:

List all executions in the Test-env pipeline

When to Use: - Quick overview of available executions - When you need execution names but not full details - Before calling cmf_show_executions for detailed information

See Also: cmf_show_executions for full execution details

cmf_execution_lineage¤

Fetches execution lineage for a specific execution UUID.

Description: Retrieves the complete lineage tree showing how an execution relates to other executions, providing a graph of execution dependencies and relationships.

Parameters:

Parameter	Type	Required	Description
`pipeline`	string	Yes	Name of the pipeline containing the execution
`selected_uuid`	string	Yes	Execution UUID (only first 4 characters are used)
`cmfClient_instances`	List[str]	No	Specific CMF Server URLs to query

Returns: Execution lineage tree showing parent-child relationships and execution flow.

Example Usage:

What is the execution lineage for UUID abc123?

Show me the lineage tree for execution abc1 in Test-env

Important Notes: - Only the first 4 characters of the UUID are used for matching - You should validate the UUID exists before calling this tool - The response contains a hierarchical tree structure

Sample Response:

[
  {
    "cmfClient": "http://server:8080",
    "data": {
      "nodes": [...],
      "edges": [...],
      "hierarchy": {...}
    }
  }
]

See Also: cmf_show_executions to find valid UUIDs, Examples: Execution Lineage

Artifact Tools¤

cmf_show_artifact_types¤

Lists all artifact types available in CMF Server(s).

Description: Retrieves the complete list of artifact types tracked in your CMF infrastructure, such as "Dataset", "Model", "Metrics", etc.

Parameters:

Parameter	Type	Required	Description
`cmfClient_instances`	List[str]	No	Specific CMF Server URLs to query

Returns: List of artifact type names.

Example Usage:

What artifact types are available?

List all artifact types in CMF

Sample Response:

[
  {
    "cmfClient": "http://server:8080",
    "data": ["Dataset", "Model", "Metrics", "Step_Metrics"]
  }
]

See Also: cmf_show_artifacts to query artifacts of a specific type

cmf_show_artifacts¤

Retrieves artifacts of a specific type for a given pipeline.

Description: Queries artifacts matching a specific type within a pipeline, returning detailed metadata for each artifact.

Parameters:

Parameter	Type	Required	Description
`pipeline`	string	Yes	Name of the pipeline to query
`artifact_type`	string	Yes	Type of artifact (e.g., "Model", "Dataset")
`cmfClient_instances`	List[str]	No	Specific CMF Server URLs to query

Returns: List of artifacts with their properties, IDs, and metadata.

Example Usage:

Show me all Model artifacts in the Test-env pipeline

Get all Datasets from the training-pipeline

Validation: - Use cmf_show_artifact_types to verify artifact type exists - Use cmf_show_pipelines to verify pipeline name

Sample Response:

[
  {
    "cmfClient": "http://server:8080",
    "data": [
      {
        "id": "42",
        "name": "trained_model.pkl",
        "type": "Model",
        "uri": "s3://bucket/models/trained_model.pkl",
        ...
      }
    ]
  }
]

See Also: cmf_show_model_card for detailed model information

cmf_artifact_lineage¤

Fetches the complete artifact lineage tree for a pipeline.

Description: Retrieves the artifact lineage graph showing how artifacts are related through executions, providing a comprehensive view of data flow through your pipeline.

Parameters:

Parameter	Type	Required	Description
`pipeline`	string	Yes	Name of the pipeline to analyze
`cmfClient_instances`	List[str]	No	Specific CMF Server URLs to query

Returns: Artifact lineage tree showing relationships between artifacts and executions.

Example Usage:

Show me the artifact lineage for Test-env

What is the artifact flow in the training-pipeline?

Important Notes: - Returns a hierarchical structure with nodes and edges - Shows which executions consume and produce which artifacts - Useful for understanding data dependencies

Sample Response:

[
  {
    "cmfClient": "http://server:8080",
    "data": {
      "nodes": [
        {"id": "artifact-1", "type": "Dataset", ...},
        {"id": "artifact-2", "type": "Model", ...}
      ],
      "edges": [
        {"source": "artifact-1", "target": "execution-1"},
        {"source": "execution-1", "target": "artifact-2"}
      ]
    }
  }
]

Additional Tools¤

cmf_show_model_card¤

Retrieves the model card for a Model artifact.

Description: Fetches comprehensive model metadata including training details, input/output artifacts, execution information, and model properties. Returns four distinct sections of model information.

Parameters:

Parameter	Type	Required	Description
`model_id`	string	Yes	ID of the model artifact (obtained from `cmf_show_artifacts`)
`cmfClient_instances`	List[str]	No	Specific CMF Server URLs to query

Returns: Four sections of model card data:

Model Data - Basic model properties and metadata
Model Execution - Training execution details and parameters
Model Input Artifacts - Datasets and artifacts used for training
Model Output Artifacts - Generated artifacts and metrics

Example Usage:

Get the model card for model ID 42

Show me training details for the model in Test-env

Validation: - Use cmf_show_artifacts with artifact_type="Model" to get valid model IDs - The model_id is returned as the id field in artifact listings

Sample Response:

[
  {
    "cmfClient": "http://server:8080",
    "data": [
      {
        "Model Data": {
          "id": "42",
          "name": "trained_model.pkl",
          "version": "1.0",
          ...
        },
        "Model Execution": {
          "execution_id": "123",
          "parameters": {...},
          ...
        },
        "Model Input Artifacts": [...],
        "Model Output Artifacts": [...]
      }
    ]
  }
]

See Also: cmf_show_artifacts to find model IDs, Examples: Model Cards

Multi-Server Queries¤

All tools support the optional cmfClient_instances parameter to query specific CMF Servers:

Query all configured servers (default):

# Parameter omitted or None
cmf_show_pipelines()

Query specific servers:

# Provide list of CMF Server URLs
cmf_show_pipelines(cmfClient_instances=["http://server1:8080", "http://server2:8080"])

Results are aggregated with each server's response clearly labeled.

Error Handling¤

Tools return error information when issues occur:

[
  {
    "cmfClient": "http://server:8080",
    "error": "Pipeline 'invalid-name' not found"
  }
]

Common Errors:

Pipeline not found - Verify pipeline name with cmf_show_pipelines
Execution UUID not found - Verify UUID with cmf_show_executions
Invalid artifact type - Check types with cmf_show_artifact_types
Connection refused - Ensure CMF Server is running and accessible

Best Practices¤

Validate inputs - Use listing tools before querying specific resources
Handle errors gracefully - Check for error fields in responses
Query specific servers - Use cmfClient_instances when you know the target server
Start broad, then narrow - List pipelines → List executions → Get lineage
Check the documentation - See Examples for real-world usage patterns

Next Steps¤

Examples - See these tools in action with real-world queries
Configuration - Learn about multi-server setup
Quick Start - Get the MCP Server running