Power Automate Copilot DevKit

Power Automate Copilot DevKit brings Power Automate cloud flow editing into VS Code, with guardrails for working safely against real Dataverse solutions.

• Edit flows like code — download them as readable JSON, edit locally, and upload safely with built-in guardrails that catch conflicts before publish to prevent us from overwriting the server copy if it has changed since you pulled it.
• Supercharge Copilot with bundled skills — turn GHCP into a Power Automate expert that authors flow JSON, expressions, error handling, and performance patterns the right way.
• Fix failing flows in minutes — let GHCP, with guided instructions, troubleshoot the failed run, root-cause the issue against the flow and fix it, then resubmit the run, all without leaving VS Code.

Features

GHCP Skills & Static Analysis

• GHCP skills help Copilot generate higher-quality Power Automate JSON, expressions, error-handling patterns, and performance-optimized designs.
• Static analysis and linting detect issues such as invalid runAfter targets, missing connection references, foreach race conditions, missing error handling, platform limits, and expression pitfalls.

Flow Lifecycle & ALM Operations

• Download flows as editable JSON under solutions/<solution-name>/Workflows/.
• Upload flows with smart validation, including server drift detection, diff review, remote backup, and connection-reference checks.
• Compare a local flow with the server copy before uploading.
• Pull a server flow and discard local changes when the cloud version should win.
• List and link connection references to support environment-specific connections.
• Refresh connection references on a flow after a connection is rebound in the environment.
• Open flows or solutions directly in the Power Automate and Power Apps maker portals.

Run Telemetry & AI-Assisted Triage

• Analyze Failed Flow Run with Copilot — download the latest failed run's action error report, save it under ref/error/, and hand it to GitHub Copilot Chat for root-cause analysis alongside the local flow JSON. Available as a per-flow tree action ($(search) icon) and as the #analyzeFailedFlowRun language-model tool.
• Resubmit Flow Run — resubmit a previously-failed run from VS Code once the underlying flow has been fixed.

Dataverse Metadata For Authoring

• Built-in metadata tools (#listDataverseTables, #dataverseTableMetadata, #dataverseOptionSet) let Copilot resolve table logical names, attribute LogicalNames, lookup binding shapes (@odata.bind), required-on-create fields, and picklist integer values directly from your environment — so it stops guessing when writing Dataverse actions.
• Per-environment on-disk cache under .power-automate-copilot-devkit/dataverse-metadata/<envId>/ with a per-call Use Cached / Refresh prompt. Clear it from the palette via Power Automate: Clear Dataverse Metadata Cache.

Installation Guide

Prerequisites

Before installing the extension, make sure you have:

• VS Code 1.90.0 or newer.
• A Microsoft account that can access your target Power Platform environment.
• Access to a Power Platform environment.
• An unmanaged Power Platform solution that contains the cloud flow you want to edit.
• Optional: GitHub Copilot recommended for natural-language flow editing and the bundled GHCP skill guidance.

Installation from VS Code Extension

Search for Power Automate Copilot and install

Usage Guide

Initial Setup

1. Click the Power Automate lightning icon in the VS Code Activity Bar.
2. Select Sign in to Power Automate... or run Power Automate: Sign In.
3. Complete the Power Platform sign-in flow.
4. Select the environment that contains your solution.
5. Select the solution that contains the flow.
6. Install the GHCP skills when prompted, or run Power Automate: Install Flow Skill into Workspace.
7. Run Download Solution (Export + Unpack) from the pinned solution.
8. Open the downloaded flow under solutions/<solution-name>/Workflows/*.json.

Editing A Flow

You can edit a flow in two ways:

• Open the flow definition from the Power Automate tree and edit the JSON manually.
• Ask GitHub Copilot to update the flow JSON using natural language.

Example: Adding an Email action

Uploading A Flow

GHCP prompts:

Upload Flow

Example:

Viewing A Flow

GHCP prompts:

View Flow

Example:

Troubleshooting A Flow

When a cloud flow run fails, you can pull its error report straight into VS Code and have GHCP root-cause it against the local flow JSON.

1. Expand the failing flow in the Power Automate tree and click the Analyze Failed Flow Run with Copilot action ($(search) icon on the flow row), or run Power Automate: Analyze Failed Flow Run with Copilot from the command palette.
2. The extension downloads the latest failed run, saves the action-level error report under ref/error/<flow>/, opens it in the editor, and hands it to GitHub Copilot Chat alongside the local flow JSON.
3. Review Copilot's root-cause analysis, then ask it to apply the fix to the flow JSON.

GHCP prompts:

Analyze the latest failed run for this flow.

After editing:

1. Save the flow file.
2. Review warnings and errors in the Problems panel.
3. Run Upload Flow when you are ready to push the local flow to the server.

Useful GHCP prompts:

Upload this flow to push my local changes to the server.
Pull this flow and discard my local changes.
View this flow in the Power Automate portal.
Download the pinned solution.
List my SharePoint connection references.
Help me add error handling to this flow.
Help me make this flow use safer retry and runAfter patterns.

Common Commands

What Upload Checks

Upload Flow updates one flow definition. It does not pack and import the entire solution.

Before uploading, the extension:

• Parses the flow JSON.
• Runs static validation and linting.
• Warns when connection references are missing or not bound to an active connection.
• Checks whether the server copy changed since the last download.
• Lets you view a diff, upload your version, pull the server version, or cancel when drift is detected.
• Backs up the current remote flow JSON under .power-automate-copilot-devkit/backups/.
• Deactivates and reactivates active flows during upload when configured to do so.

GHCP Skill Docs

The extension includes Power Automate guidance for GitHub Copilot under resources/skill/docs/flow-skill/. Installing the skill into your workspace helps Copilot follow the project's flow-authoring patterns.

The bundled docs cover:

• Error handling and try/catch/finally scope patterns.
• Retry, runAfter, and failure-path guidance.
• Performance and platform limits.
• Expert flow patterns.
• Power Automate expression examples.

User Settings

Most users can keep the defaults.

In untrusted workspaces, workspace-scoped solutionsRoot values are ignored. Use user or global settings, or trust the workspace.

Flow APIs (advanced, opt-in)

Most features in this extension talk to Dataverse (https://<org>.crm*.dynamics.com) using VS Code's built-in Microsoft account provider. Those work out of the box with no extra setup.

A small set of capabilities — environment auto-discovery via api.flow.microsoft.com, failed-run inspection, run resubmit, and a few connection-reference helpers — hit the Power Automate Service API instead. VS Code's built-in first-party client is not pre-authorized for that resource and returns AADSTS65002 on the token request, so the extension does not call the Flow API unless you opt in.

To opt in:

1. Register an Entra (Azure AD) app in your tenant (multi-tenant, public-client, with http://localhost and https://vscode.dev/redirect as redirect URIs).
2. Grant it delegated permissions: Power Automate Service User, Dataverse user_impersonation, Microsoft Graph User.Read.
3. Grant tenant-wide admin consent for those permissions. This step is non-optional — without admin consent the Flow API call still fails, and home-tenant users may additionally hit AADSTS650051 ("SPN already present") on first sign-in.
4. Run Power Automate: Bring Your Own AAD App Registration (Advanced) and paste the app's Client ID and Tenant ID, or set powerAutomateCopilotDevKit.aadClientId / powerAutomateCopilotDevKit.aadTenantId directly.
5. If your tenant has never used Power Automate, also run Power Automate: Grant Power Automate Access once to provision the Power Automate Service principal in your directory.

Until admin consent has been granted, the extension will fall back to manual environment URL entry and skip the Flow API code paths — every Dataverse-only feature (download, upload, drift check, validation, GHCP skills, Dataverse metadata tools) keeps working without it.

Troubleshooting

Developer Guide

Use this section when you want to build, debug, package, or extend the extension code.

Project Structure

How To Extend

Add A Command

1. Add the command contribution in package.json.
2. Add menu placement in package.json if the command should appear in the tree, editor, Explorer, or command palette.
3. Implement the command under src/commands/.
4. Register the command in src/extension.ts with the existing register(...) helper.
5. Reuse existing services such as AuthService, DataverseClient, and FlowTreeProvider instead of creating parallel access paths.

Add A GHCP Tool

1. Add contributes.languageModelTools metadata in package.json.
2. Implement the tool under src/tools/.
3. Register the tool in src/extension.ts with vscode.lm.registerTool.
4. Prefer wrapping an existing command/service path so user confirmations and safeguards stay consistent.

Add Validation

1. Add rules in src/validation/flowLinter.ts.
2. Keep findings read-only; validation should not rewrite the user's flow JSON.
3. Return stable rule IDs, severities, JSON paths, and offsets when possible.
4. Update diagnostics or tests if validation behavior changes.

Add Tree Behavior

1. Update src/tree/FlowTreeProvider.ts for new nodes, labels, icons, or refresh behavior.
2. Add or adjust contextValue strings for menu targeting.
3. Update package.json menu contributions to show commands in the right tree context.
4. Keep network work lazy and cached where possible so the tree stays responsive.

Add Configuration

1. Add the setting under contributes.configuration.properties in package.json.
2. Read values through src/config.ts.
3. Use trusted configuration helpers for settings that affect local executable paths or workspace-relative paths.
4. Document user-facing settings in the User Guide.

Add Or Update GHCP Skill Docs

1. Update files under resources/skill/docs/flow-skill/.
2. Keep guidance concise and directly useful for editing Power Automate flow JSON.
3. Confirm Power Automate: Install Flow Skill into Workspace still installs the expected files.

Safety And Design Notes

• Validate solution names, workflow IDs, and environment IDs before using them in CLI or Dataverse operations.
• Preserve upload safeguards: linting, connection checks, drift detection, remote backup, ETag conditional PATCH, and state restore behavior.
• Keep upload scoped to one flow unless the product intentionally changes.
• Respect untrusted workspace behavior for solutionsRoot.
• Prefer existing command and service paths for GHCP tools so agent-mode behavior matches the UI.
• Do not bypass FlowManifest when changing download, upload, drift, backup, or baseline behavior.