ReferenceRoles and permissions

MLflow

Experiments, runs, models, traces, workspaces and webhooks.

Agentic Friendly

AI Viewer and AI Builder are MLflow functional roles. AI Viewer grants experiment, run, registry and trace reads; AI Builder includes those reads and the supported writes. Tenant administration alone does not grant those functional permissions.

Prompts and logged models use mlflow.model.*. Evaluation datasets use mlflow.experiment.*. Some compound operations require two permissions, such as linking a prompt version to a run (mlflow.model.update and mlflow.run.update).

Workspace and webhook permissions use a different rule: membership supplies the operation gate, while workspace access decides the result. Current members can read their tenant's workspaces. The workspace owner, tenant administrator or platform administrator can manage the workspace and its webhooks, subject to the API's tenant-context checks. Creating a workspace requires mlflow.workspace.create.

Workspace management does not redirect experiment, run, registry or trace APIs into that workspace. Those APIs use MLflow's default workspace; webhooks are workspace-aware. The permissions listed here are not a claim that every MLflow entity is isolated by workspace.

Deleting experiments and runs is reversible through their restore operations, which require the corresponding delete permission. Registry deletion is permanent. Model promotion changes stages or aliases; runtime deployment requires the separate platform permission.

Permissions

This family contains 30 permissions. “Baseline member” means every current tenant member already has the operation permission, including members with any system role. It does not bypass the resource checks described above. “Custom role” means the identifier can be selected when defining a custom role.

mlflow.experiment.create

Create an MLflow experiment or an associated evaluation dataset through the supported API.

Create an MLflow experiment or an associated evaluation dataset through the supported API.

PropertyValue
ScopeSelected tenant
Granted byAI Builder
Custom roleYes
Also impliesmlflow.experiment.list, mlflow.experiment.read

mlflow.experiment.delete

Soft-delete or restore an experiment, or delete an associated evaluation dataset through its API.

Soft-delete or restore an experiment, or delete an associated evaluation dataset through its API. Dataset deletion is distinct from experiment soft deletion.

PropertyValue
ScopeSelected tenant
Granted byAI Builder
Custom roleYes
Also impliesmlflow.experiment.list, mlflow.experiment.read, mlflow.experiment.update

mlflow.experiment.list

List or search visible MLflow experiments.

List or search visible MLflow experiments.

PropertyValue
ScopeSelected tenant
Granted byAI Viewer, AI Builder
Custom roleYes
Also impliesmlflow.experiment.read

mlflow.experiment.read

Read an experiment and its supported evaluation-dataset information.

Read an experiment and its supported evaluation-dataset information.

PropertyValue
ScopeSelected tenant
Granted byAI Viewer, AI Builder
Custom roleYes

mlflow.experiment.update

Change experiment metadata or associated evaluation-dataset tags and experiment associations.

Change experiment metadata or associated evaluation-dataset tags and experiment associations.

PropertyValue
ScopeSelected tenant
Granted byAI Builder
Custom roleYes
Also impliesmlflow.experiment.list, mlflow.experiment.read

mlflow.model.create

Create registered models or versions, prompt versions, or logged models through supported APIs.

Create registered models or versions, prompt versions, or logged models through supported APIs.

PropertyValue
ScopeSelected tenant
Granted byAI Builder
Custom roleYes
Also impliesmlflow.model.list, mlflow.model.read

mlflow.model.delete

Permanently delete registry objects or versions through supported APIs.

Permanently delete registry objects or versions through supported APIs. This is not experiment or run soft deletion.

PropertyValue
ScopeSelected tenant
Granted byAI Builder
Custom roleYes
Also impliesmlflow.model.list, mlflow.model.read, mlflow.model.update

mlflow.model.list

List or search registered models, prompt registry entries and logged models through supported APIs.

List or search registered models, prompt registry entries and logged models through supported APIs.

PropertyValue
ScopeSelected tenant
Granted byAI Viewer, AI Builder
Custom roleYes
Also impliesmlflow.model.read

mlflow.model.promote

Manage model-version stages or aliases and supported prompt aliases.

Manage model-version stages or aliases and supported prompt aliases. This does not deploy a model to the inference runtime.

PropertyValue
ScopeSelected tenant
Granted byAI Builder
Custom roleYes
Also impliesmlflow.model.list, mlflow.model.read

mlflow.model.read

Read registry objects, their versions and supported metadata.

Read registry objects, their versions and supported metadata.

PropertyValue
ScopeSelected tenant
Granted byAI Viewer, AI Builder
Custom roleYes

mlflow.model.update

Change registry metadata, tags or supported logged-model state.

Change registry metadata, tags or supported logged-model state.

PropertyValue
ScopeSelected tenant
Granted byAI Builder
Custom roleYes
Also impliesmlflow.model.list, mlflow.model.read

mlflow.run.create

Create a run in an accessible experiment.

Create a run in an accessible experiment.

PropertyValue
ScopeSelected tenant
Granted byAI Builder
Custom roleYes
Also impliesmlflow.run.list, mlflow.run.read

mlflow.run.delete

Soft-delete or restore an MLflow run.

Soft-delete or restore an MLflow run.

PropertyValue
ScopeSelected tenant
Granted byAI Builder
Custom roleYes
Also impliesmlflow.run.list, mlflow.run.read, mlflow.run.update

mlflow.run.list

List or search runs in visible experiments.

List or search runs in visible experiments.

PropertyValue
ScopeSelected tenant
Granted byAI Viewer, AI Builder
Custom roleYes
Also impliesmlflow.run.read

mlflow.run.read

Inspect a run's metrics, parameters, tags and accessible artifacts.

Inspect a run's metrics, parameters, tags and accessible artifacts.

PropertyValue
ScopeSelected tenant
Granted byAI Viewer, AI Builder
Custom roleYes

mlflow.run.update

Change run metadata or lifecycle state, log parameters and metrics, or upload artifacts through supported operations.

Change run metadata or lifecycle state, log parameters and metrics, or upload artifacts through supported operations.

PropertyValue
ScopeSelected tenant
Granted byAI Builder
Custom roleYes
Also impliesmlflow.run.list, mlflow.run.read

mlflow.trace.delete

Delete individual traces or a selected batch of traces.

Delete individual traces or a selected batch of traces.

PropertyValue
ScopeSelected tenant
Granted byAI Builder
Custom roleYes
Also impliesmlflow.trace.list, mlflow.trace.read, mlflow.trace.update

mlflow.trace.list

Search traces in accessible experiments.

Search traces in accessible experiments.

PropertyValue
ScopeSelected tenant
Granted byAI Viewer, AI Builder
Custom roleYes
Also impliesmlflow.trace.read

mlflow.trace.read

Read trace details and trace assessments.

Read trace details and trace assessments.

PropertyValue
ScopeSelected tenant
Granted byAI Viewer, AI Builder
Custom roleYes

mlflow.trace.update

Change trace tags, link traces to runs, or create, update and delete trace assessments.

Change trace tags, link traces to runs, or create, update and delete trace assessments.

PropertyValue
ScopeSelected tenant
Granted byAI Builder
Custom roleYes
Also impliesmlflow.trace.list, mlflow.trace.read

mlflow.webhook.create

Create a webhook in an administrable workspace.

Create a webhook in an administrable workspace.

PropertyValue
ScopeSelected tenant
Granted byBaseline member
Custom roleNo. Baseline tenant membership; access is decided on the individual MLflow workspace.

mlflow.webhook.delete

Delete a webhook in an administrable workspace.

Delete a webhook in an administrable workspace.

PropertyValue
ScopeSelected tenant
Granted byBaseline member
Custom roleNo. Baseline tenant membership; access is decided on the individual MLflow workspace.

mlflow.webhook.list

List webhooks in a workspace the caller is authorized to administer.

List webhooks in a workspace the caller is authorized to administer.

PropertyValue
ScopeSelected tenant
Granted byBaseline member
Custom roleNo. Baseline tenant membership; access is decided on the individual MLflow workspace.
Also impliesmlflow.webhook.read

mlflow.webhook.read

Read a webhook in an administrable workspace.

Read a webhook in an administrable workspace.

PropertyValue
ScopeSelected tenant
Granted byBaseline member
Custom roleNo. Baseline tenant membership; access is decided on the individual MLflow workspace.

mlflow.webhook.update

Change or test a webhook in an administrable workspace.

Change or test a webhook in an administrable workspace.

PropertyValue
ScopeSelected tenant
Granted byBaseline member
Custom roleNo. Baseline tenant membership; access is decided on the individual MLflow workspace.

mlflow.workspace.create

Create an MLflow workspace associated with the active tenant.

Create an MLflow workspace associated with the active tenant. The creator receives ownership.

PropertyValue
ScopeSelected tenant
Granted byTenant administrator
Custom roleYes

mlflow.workspace.delete

Delete an MLflow workspace only when the caller has workspace-level delete authority.

Delete an MLflow workspace only when the caller has workspace-level delete authority.

PropertyValue
ScopeSelected tenant
Granted byBaseline member
Custom roleNo. Baseline tenant membership; access is decided on the individual MLflow workspace.

mlflow.workspace.list

List MLflow workspaces the caller can access.

List MLflow workspaces the caller can access.

PropertyValue
ScopeSelected tenant
Granted byBaseline member
Custom roleNo. Baseline tenant membership; access is decided on the individual MLflow workspace.
Also impliesmlflow.workspace.read

mlflow.workspace.read

Read an accessible MLflow workspace.

Read an accessible MLflow workspace.

PropertyValue
ScopeSelected tenant
Granted byBaseline member
Custom roleNo. Baseline tenant membership; access is decided on the individual MLflow workspace.

mlflow.workspace.update

Update an MLflow workspace only when the caller has workspace-level update authority.

Update an MLflow workspace only when the caller has workspace-level update authority.

PropertyValue
ScopeSelected tenant
Granted byBaseline member
Custom roleNo. Baseline tenant membership; access is decided on the individual MLflow workspace.

On this page