Powerbi Model Edit

/powerbi-model-edit — Live Edit a Power BI Model via MCP

You are editing a running Power BI Desktop model via the powerbi MCP server. Operations are interactive: measures, tables, DAX validation, TMDL export.

This skill is local-only — it does not touch Fabric. To publish to the service after editing, use /powerbi-publish.

Prerequisites

• Power BI Desktop running with a model open (.pbix or .SemanticModel)
• powerbi MCP server configured in .mcp.json (see ONBOARDING.md Step 7)
• Model must be fully loaded (not refreshing)

Usage

/powerbi-model-edit                                       # connect, show inventory
/powerbi-model-edit measures                              # list all measures
/powerbi-model-edit add-measure {table} {name} {dax}      # add a measure
/powerbi-model-edit tables                                # list tables
/powerbi-model-edit validate "{dax_expression}"           # validate DAX
/powerbi-model-edit export {output_dir}                   # export model to TMDL folder

Behavior

Step 1: Discover instances

mcp__powerbi__connection_operations(operation="ListLocalInstances")

If multiple: ask the user. If none: instruct to open the model in PBI Desktop.

Step 2: Connect

mcp__powerbi__connection_operations(operation="Connect", dataSource="localhost:{port}")

Step 3: Execute operation

measures — List

mcp__powerbi__measure_operations(operation="List")

Render: name, table, format string, expression preview.

add-measure {table} {name} {dax}

Validate first:

mcp__powerbi__dax_query_operations(
  operation="Validate",
  query="EVALUATE {{ {dax} }}"
)

If valid, create:

mcp__powerbi__measure_operations(
  operation="Create",
  definitions=[{
    tableName: "{table}",
    name: "{name}",
    expression: "{dax}",
    formatString: "#,##0.00"
  }]
)

Remind the user: changes only persist if they save in Power BI Desktop.

tables — List

mcp__powerbi__table_operations(operation="List")

validate "{dax}"

mcp__powerbi__dax_query_operations(operation="Validate", query="EVALUATE {dax}")

Show errors with line/column info if any.

export {output_dir}

mcp__powerbi__database_operations(
  operation="ExportToTmdlFolder",
  tmdlFolderPath="{output_dir}/{ModelName}.SemanticModel"
)

Use this regularly for version control — TMDL is text-friendly under git.

MCP Operations Reference

When to Use This vs /powerbi-model-create

Logging

Read-only operations (measures, tables, validate, export) are not logged to ops.log.

Write operations (add-measure, future edit/delete):

{ISO_timestamp} | {role} | PBI-EDIT | local | {operation}: {detail} | {ok|fail}

Troubleshooting

• No instances found → open the model in PBI Desktop, wait for full load
• Connection timeout → restart PBI Desktop, ensure MCP server is running
• Repeated MCP failures → connection_operations(operation="Disconnect"), restart PBI Desktop, reconnect

Notes

• Changes via MCP require a manual save in Power BI Desktop to persist on disk.
• Export to TMDL regularly so your changes survive in git.
• For complex DAX, run Execute (not just Validate) to confirm result.

Visual verification of bound reports

DAX edits via MCP only mutate the local PBI Desktop session. Once persisted on disk (manual save) and published via /powerbi-publish, downstream reports that bind to this model may behave differently — a renamed measure breaks visuals that reference it by name, a changed format string changes display formatting, a relationship edit can change row counts.

After publishing the edited model, use the Playwright visual-feedback loop to verify at least one consuming report still renders correctly:

• For reports: /<client>-pbi-loop verify {report-ref} --env {env}
• For other distributions: see core/playbooks/playwright-pbi-loop.md for the general pattern

This is especially important for: measure renames, removed measures, format-string changes, relationship modifications. Pure additive measure work (new measure, no existing visual touched) is lower risk.