Purpose-built Microsoft Fabric scaffolding for AI coding agents. This starter helps external developers and teams move faster with pre-configured Fabric skills, reusable prompts, and agent instructions for GitHub Copilot, Claude Code, Cursor, Windsurf, and Codex.
It is designed for practical Fabric work: Lakehouse engineering, Warehouse and SQL endpoint development, Eventhouse and KQL workflows, Power BI semantic model operations, REST API integration, and cross-workload orchestration.
- Fabric-first structure instead of a generic AI agent template
- High-value Fabric skills for Lakehouse, Warehouse, Eventhouse, Power BI, and REST APIs
- Shared instructions that keep multiple coding agents aligned on the same platform conventions
- Ready-to-use MCP setup assets for Fabric-focused agent workflows
- Built-in guidance for auth scopes, pagination, long-running operations, throttling, and OneLake patterns
- Python 3.13+ (see
.python-version) - Azure CLI authenticated (
az login) - Microsoft Fabric capacity (F2+ or P1+)
- Node.js 18+ (optional, for Fabric MCP server)
az login
az account get-access-token --resource https://api.fabric.microsoft.com| Agent | Entry Point |
|---|---|
| GitHub Copilot | .github/copilot-instructions.md → auto-loaded |
| Claude Code | CLAUDE.md → auto-loaded |
| Cursor | .cursorrules → auto-loaded |
| Windsurf | .windsurfrules → auto-loaded |
| Codex / Jules | AGENTS.md → auto-loaded |
| All agents | .agents/instructions.md → full reference |
All entry points route to .agents/instructions.md for shared skills and Fabric patterns.
If you need to register a remote Fabric MCP server with GitHub Copilot CLI, Claude Desktop, or VS Code, use the tracked setup assets in mcp-setup/.
Examples:
.\mcp-setup\register-fabric-mcp.ps1 -ServerUrl "https://your-fabric-mcp-server.com" -ServerName "fabric"./mcp-setup/register-fabric-mcp.sh --server-url "https://your-fabric-mcp-server.com" --server-name "fabric"Use mcp-setup/mcp-config-template.json as a starting point for manual configuration.
All skills are under .agents/skills/. Each has a SKILL.md with YAML frontmatter and structured guidance.
| Skill | Scope |
|---|---|
fabric-core |
Topology, auth scopes, REST API, workspace/item resolution, pagination, LRO, throttling, OneLake |
fabric-lakehouse |
Lakehouse design, Delta tables, schemas, shortcuts, security, V-Order, PySpark, medallion |
fabric-api-discovery |
Fabric MCP Server tools, API specs, item definitions, developer vs consumer patterns |
fabric-alm-cicd |
Git integration, deployment pipelines, Variable Libraries, REST/fabric-cicd automation, release validation |
fabric-data-agent |
Fabric Data Agent design, source routing, instructions, example queries, evaluation, governance |
microsoft-code-reference |
Verify Microsoft SDK methods, packages, and parameter signatures against official docs |
Endpoint Skills (from microsoft/skills-for-fabric v0.3.1)
| Skill | Purpose |
|---|---|
spark-authoring-cli |
Spark and Data Engineering workflows |
spark-consumption-cli |
Interactive Lakehouse table analysis |
spark-operations-cli |
Spark job, session, and performance diagnostics |
sqldw-authoring-cli |
Warehouse, SQL Endpoint authoring |
sqldw-consumption-cli |
Read-only SQL queries |
sqldw-operations-cli |
Warehouse performance diagnostics and query insights |
eventhouse-authoring-cli |
KQL table management, ingestion, policies |
eventhouse-consumption-cli |
Read-only KQL queries |
eventstream-authoring-cli |
Eventstream topology authoring and deployment |
eventstream-consumption-cli |
Eventstream inspection and monitoring |
dataflows-authoring-cli |
Dataflows Gen2 authoring and Power Query M definitions |
dataflows-consumption-cli |
Dataflows Gen2 read-only exploration and monitoring |
dataflows-save-as-authoring-cli |
Dataflows Gen1 to Gen2 save-as upgrade workflows |
activator-authoring-cli |
Activator/Reflex alert and action authoring |
activator-consumption-cli |
Activator/Reflex definition inspection |
powerbi-authoring-cli |
Semantic model creation, TMDL, refresh |
powerbi-consumption-cli |
DAX queries and metadata |
search-consumption-cli |
Fabric catalog search and item discovery |
databricks-migration |
Databricks to Fabric migration planning |
synapse-migration |
Azure Synapse to Fabric migration |
hdinsight-migration |
Azure HDInsight to Fabric migration |
e2e-medallion-architecture |
Bronze/Silver/Gold lakehouse patterns |
| Agent | Scope |
|---|---|
FabricDataEngineer |
Cross-workload orchestration: medallion, ETL/ELT, migration, data quality |
FabricSolutionArchitect |
Cross-workload architecture, workload selection, trade-offs, governance, ALM strategy |
FabricReleaseEngineer |
Release engineering across Git integration, deployment pipelines, Variable Libraries, CI/CD gates |
FabricQualityReviewer |
Fabric code, definition, agent, skill, and release-readiness review |
FabricAdmin |
Capacity, governance, security, cost |
FabricAppDev |
Application development with Fabric |
FabricMigrationEngineer |
Migration planning for Synapse, Databricks, and HDInsight to Fabric |
| File | Purpose |
|---|---|
COMMON-CORE.md |
Fabric topology, auth, scopes, REST API patterns, pagination, LRO, throttling |
COMMON-CLI.md |
CLI recipes (az rest, pagination handling, job execution) |
FABRIC-ALM-CORE.md |
Fabric Git integration, deployment pipelines, Variable Libraries, CI/CD validation, rollback |
FABRIC-DATA-AGENT-CORE.md |
Fabric Data Agent source selection, instructions, examples, evaluation, governance, ALM |
ITEM-DEFINITIONS-CORE.md |
Item definition envelope, create/update/export/import |
SPARK-AUTHORING-CORE.md |
Spark/Livy/Delta table patterns |
SPARK-CONSUMPTION-CORE.md |
Spark read/query and Livy consumption patterns |
SPARK-MONITORING-CORE.md |
Spark monitoring API and diagnostics patterns |
SPARK-NOTEBOOK-AUTHORING-CORE.md |
Notebook authoring and lakehouse context guidance |
SQLDW-AUTHORING-CORE.md |
T-SQL surface area and limitations |
SQLDW-CONSUMPTION-CORE.md |
SQL read/query, security, metadata refresh, and performance patterns |
DATAFLOWS-AUTHORING-CORE.md |
Dataflows Gen2 definition and authoring patterns |
DATAFLOWS-CONSUMPTION-CORE.md |
Dataflows Gen2 inspection and monitoring patterns |
EVENTHOUSE-AUTHORING-CORE.md |
KQL ingestion and policy patterns |
EVENTHOUSE-CONSUMPTION-CORE.md |
KQL query, Eventhouse discovery, and analytics patterns |
EVENTSTREAM-AUTHORING-CORE.md |
Eventstream topology authoring patterns |
EVENTSTREAM-CONSUMPTION-CORE.md |
Eventstream inspection and monitoring patterns |
Base URL: https://api.fabric.microsoft.com/v1/
| Target | Scope |
|---|---|
| Fabric REST API | https://api.fabric.microsoft.com/.default |
| OneLake (DFS/Blob) | https://storage.azure.com/.default |
| Warehouse / SQL Endpoint | https://database.windows.net/.default |
| KQL / Kusto | https://kusto.kusto.windows.net/.default |
| Power BI / XMLA | https://analysis.windows.net/powerbi/api/.default |
- Pagination: List APIs return
continuationToken. Repeat with?continuationToken=<value>until null - LRO: Mutating ops may return
202withLocation,x-ms-operation-id,Retry-After. Poll untilSucceededorFailed - Throttling:
429includesRetry-After. Must wait before retrying
| Tool | Install | Purpose |
|---|---|---|
| Fabric CLI | pip install ms-fabric-cli |
Command-line Fabric operations |
| fabric-cicd | pip install fabric-cicd |
CI/CD automation |
| Fabric Pro-Dev MCP | npm install -g fabric-pro-dev-mcp-server |
Local MCP server for API discovery |
This starter now includes additional web-researched agent patterns from Microsoft Fabric ALM guidance, Fabric Data Agent guidance, VS Code Copilot customization docs, and the community awesome-copilot pattern of separating agents, skills, prompts, and review workflows.
| Workflow | Entry Point | Use For |
|---|---|---|
| Fabric release plan | .agents/prompts/plan-fabric-release.prompt.md |
Git integration, deployment pipelines, Variable Libraries, CI/CD gates |
| Data Agent design | .agents/prompts/design-data-agent.prompt.md |
Governed conversational analytics over Fabric sources |
| Fabric quality review | .agents/prompts/review-fabric-solution.prompt.md |
Review code, item definitions, agents/skills, and release readiness |
These prompts route into .agents/skills/fabric-alm-cicd/, .agents/skills/fabric-data-agent/, and .agents/agents/FabricQualityReviewer.agent.md so common tasks can be invoked without restating the full workflow each time.