Fabric Data Agent — Plugin Marketplace

Create, configure, test, tune, and publish Microsoft Fabric Data Agents from VS Code using natural language ✨

Extend the power of GitHub Copilot with the Fabric Data Agent plugin — an MCP server, skills, and agents that connect AI assistants to Microsoft Fabric's Data Agent lifecycle.

---

📋 Prerequisites

Before getting started, ensure you have:

• Azure CLI — Install, then az login
• uv — Install (Python package runner)
• Fabric workspace access (Contributor role)

az login --tenant <your-tenant-id>
az account get-access-token --resource https://api.fabric.microsoft.com

💡 Why uv? The MCP server is a Python package installed and run via uvx. No manual pip install or virtual environment needed.

---

🚀 Quick Start with GitHub Copilot CLI

# 1. Open GitHub Copilot CLI
copilot

# 2. Add this plugin marketplace (one-time setup)
/plugin marketplace add harigouthami/fabric-copilot-plugins

# 3. Install the plugin
/plugin install fabric-data-agent@fabric-copilot-plugins

That's it! Restart Copilot and start using the plugin:

You: Create a data agent called ADOWIA in A3PInsights workspace
You: Connect the External lakehouse, select tca_adowia tables
You: Run accuracy test [attach CSV]

---

📦 Alternative: Standalone MCP Installation

Add this to .vscode/mcp.json in any workspace:

{
  "servers": {
    "fabric-data-agent": {
      "command": "uvx",
      "args": [
        "--from", "git+https://github.com/harigouthami/fabric-copilot-plugins.git#subdirectory=fabric-data-agent-mcp",
        "fabric-data-agent-mcp"
      ]
    }
  }
}

Then reload VS Code (Ctrl+Shift+P → Reload Window). The 20 MCP tools will appear in Copilot Chat.

---

🎯 What You Can Do

Capability	Example
Create agents	"Create a data agent called ADOWIA in A3PInsights"
Connect data	"Connect External lakehouse, select the adowia tables"
Auto-generate instructions	"Use this ADO repo with TMDL files for knowledge"
Add few-shots	"Generate few-shots and validate SQL before adding"
Test accuracy	"Run accuracy test" (attach CSV)
Tune failures	"The agent gives wrong results for total refinements"
Publish	"Publish the agent"

---

🔌 Available Plugins

Plugin	Description
fabric-data-agent	Full lifecycle management of Fabric Data Agents — create, configure, test, tune, and publish through natural language.

---

🧩 Skills

Skill	Description
create-agent	Guided end-to-end agent setup with SQL validation
test-agent	CSV-based accuracy testing with tolerance matching
tune-agent	Diagnose and fix failing queries

---

🤖 Agents

Agent	Description
Fabric Manager	Full lifecycle: create → configure → publish → query
Data Explorer	Read-only: browse workspaces, lakehouses, tables
Agent Tester	Test & tune: CSV accuracy tests, diagnose SQL, fix few-shots

---

🔐 Authentication

All Fabric operations require Azure AD authentication:

az login
az account get-access-token --resource https://api.fabric.microsoft.com

No secrets or tokens are stored by the plugin. Uses az login — tokens auto-cached for ~60 min.

---

🤝 Contributing

We welcome new plugins! In short:

1. Create your plugin under plugins/{your-plugin}/
2. Add .mcp.json, README.md, and skills/{name}/SKILL.md
3. Register it in .github/plugin/marketplace.json
4. Submit a pull request

---

📚 Repository Structure

FND-A3P-Common-Plugins/
├── .github/plugin/marketplace.json
├── plugins/fabric-data-agent/
│   ├── .github/plugin/plugin.json
│   ├── .mcp.json
│   ├── README.md
│   ├── agents/
│   └── skills/
├── fabric-data-agent-mcp/          # MCP server source
├── hooks/
├── instructions/
└── README.md

---

📄 License

This project is licensed under the MIT License.