Agent
A conversational AI assistant that reads, writes, and modifies code in your project. It makes decisions and invokes tools to do what you ask. Examples: Claude Code, Codex, GitHub Copilot, Kilo Code.
Plugin
BI SUPERPOWERS
Your AI assistant ships Power BI work that holds up in production. 6 specialist skills, 2 official Microsoft MCP servers, one CLI.
all-agents terminal
$ npm install -g @luquimbo/bi-superpowers$ super install --all --yesRestart your agent once after install. That's the whole setup.Full install guide →
Fundamentals
New to AI agents?
Start here.
Four concepts you need. If you already work with AI agents, skip ahead.
Plugin
An installable package that gives an agent new capabilities: specialist skills, MCP connections, configuration. bi-superpowers is a plugin: one command and your agent knows Power BI.
Skill
A focused expert inside a plugin. Each skill owns one thing (modeling, DAX, performance) with its own scope, prerequisites, and handoffs. You invoke it by name, e.g. /bi-dax.
MCP
Model Context Protocol. An open standard that lets an agent talk to external tools (Power BI Desktop, a database, an API) over a local connection. The agent reads and writes through the MCP without you wiring it by hand.
Skills
6 specialists,
one cognitive contract.
Each skill is a focused expert. One job, done well. Grouped by what you're trying to do today: get started, build the model, or connect to Desktop.
Get Started
Open a session, route to the right skill, and bootstrap a brand-new BI project.
/bi-startSession OrchestratorYour starting point. Always.
Opens a clean session, checks for plugin updates, shows the skill menu, and routes the conversation to the right specialist.
Use it when
- Starting a new session
- You don't know which skill to invoke
- You want an update check before working
Open detail panel
/bi-kickoffProject AnalystFrom zero to a working BI plan.
Inspects an existing repo or scaffolds a new one, identifies gaps, and produces a sensible plan with an AGENTS.md baseline.
Use it when
- Opening a Power BI repo for the first time
- Starting a new PBIP project from scratch
- Asking the agent to understand structure before editing
Open detail panel
Build & Optimize
Design semantic models, author DAX, and squeeze every millisecond out of your reports.
/bi-modelingModel ArchitectStar schemas, relationships, TMDL.
Designs and audits semantic models. Marked Date Tables, Auto Date/Time off, calculation groups, IBCS theme. The patterns that make a model maintainable.
Use it when
- Designing a new semantic model
- Auditing an existing model for smells
- Refactoring relationships and table structure
Open detail panel
/bi-daxDAX ExpertMeasures that compute the right thing.
Authors and optimizes DAX measures, calculation groups, time intelligence, and DAX UDFs. With DAXLIB and Microsoft Learn as live references.
Use it when
- Writing a new measure or UDF
- Refactoring a slow or wrong measure
- Implementing time intelligence the right way
Open detail panel
/bi-performancePerformance EngineerDiagnose slow models. Fix them.
Profiles models with DAX Studio and Best Practice Analyzer, finds bottlenecks in storage, query plans, and refresh paths, and proposes concrete fixes.
Use it when
- A report or refresh feels slow
- You need a model audit with metrics
- You're shipping to production and want a check
Open detail panel
Connect
Talk to Power BI Desktop directly through the official Microsoft MCP server.
Template
Start from a real
PBIP template.
Every skill references the same canonical sales template. A complete PBIP package with AGENTS.md, semantic model, and report definition. Not a snippet, not a prompt. A working project you can fork.
Complete PBIP package
Ships a real Power BI project: TMDL semantic model, PBIR report definition, dataset settings, and an AGENTS.md baseline. Open it in Power BI Desktop and it just works.
Canonical reference for every skill
/bi-modeling, /bi-dax, /bi-performance. Every specialist consults the same template. Patterns stay consistent across sessions, not regenerated each time.
Sales domain, IBCS-ready
Realistic measures, dimensions, and a calculation group. Marked Date Table, Auto Date/Time off, IBCS theme. The patterns that read like a senior modeler wrote them.
Create from template via /bi-kickoff
/bi-kickoff can copy templates/sales/ into your repo as pbip-files/<projectName>.* with everything renamed. From a clean repo to a running PBIP in one command.
templates/sales/
templates/sales/
├─ AGENTS.md
├─ pbip-files/
│ ├─ Sales.SemanticModel/
│ │ ├─ definition/
│ │ │ ├─ tables/*.tmdl
│ │ │ ├─ relationships.tmdl
│ │ │ └─ cultures/
│ │ └─ model.tmdl
│ └─ Sales.Report/
│ └─ definition/
│ ├─ pages/
│ ├─ bookmarks/
│ └─ report.json
└─ Sales.pbipWorkflow
From empty repo
to shippable model.
A predictable arc through the skills. Each handoff is explicit; nothing magical happens behind your back.
- 01
/bi-startOpen the session
Update check, skill menu, environment hints. Routes you to the right specialist.
- 02
/bi-kickoffMap the project
Inspects the repo, finds gaps, writes an AGENTS.md plan you can actually follow.
- 03
/bi-connectConnect to Power BI Desktop
Wires the agent to the open model via the Microsoft Power BI Modeling MCP.
- 04
/bi-modeling · /bi-daxBuild & optimize
Design the semantic model. Author DAX. Refactor what's wrong without breaking what's right.
- 05
/bi-performanceProfile & ship
DAX Studio + Best Practice Analyzer in the loop. Ship with numbers, not hopes.
External tools
Built on
official + open-source tools.
We don't reinvent the wheel. bi-superpowers stands on Microsoft's official tooling and a battle-tested open-source editor. Three pieces, all transparent.
Microsoft · official
Power BI Modeling MCP
Local stdio · Windows
Microsoft's official MCP server that exposes the semantic model open in Power BI Desktop over XMLA. Lists tables, columns, measures, and relationships; runs DAX; edits the model.
Used by /bi-connect, /bi-modeling, /bi-dax, and /bi-performance.
View on GitHub →Microsoft · official
Microsoft Learn MCP
HTTP · cross-platform
Microsoft's official MCP server that gives the agent access to Microsoft Learn documentation in context. DAX, Power BI, Fabric. Always the current version, not the version baked into the agent's training data.
Used cross-cutting whenever a skill needs fresh docs.
View documentation →Open-source · MIT
Tabular Editor 2 CLI
CLI · Windows
Tabular model editor. The de facto standard in the Power BI community. We use it in CLI mode to run the Best Practice Analyzer (~50 official Microsoft rules) against the live Desktop model. Installed only with explicit user consent (~30 MB).
Used by /bi-performance for deep BPA audits.
View on GitHub →Under the Hood
Reproducible.
Local. Open source.
Not a loose prompt pack. A reproducible Power BI agent layer that ships the same operating contract to every supported agent.
6 specialist skills
Each skill is a focused operator with a clear scope, prerequisites, and handoffs. No vague mega-prompts.
skills/
├─ bi-start/
├─ bi-kickoff/
├─ bi-modeling/
├─ bi-dax/
├─ bi-performance/
└─ bi-connect/2 official Microsoft MCPs
Power BI Modeling MCP for live model access. Microsoft Learn MCP for current docs in context. Both first-party.
powerbi-modeling-mcp stdio
microsoft-learn http1 cross-agent CLI
super installs skills + MCPs at the right paths for each agent. No hand-editing of config files.
super install --all --yes
super upgrade
super powersNo SaaS backend
Skills are local files. MCP servers run on your machine. PBIP files stay in your repo. There is no bi-superpowers cloud. By design.
Codex-first, agent-fair
End-to-end validated in Codex. Compatibility adapters for Claude Code, GitHub Copilot, and Kilo Code. Same skills, same MCPs.
FAQ
Frequently
asked.
Does this replace Power BI Desktop?
No. Power BI Desktop stays the visual and report-authoring tool. The plugin gives the agent inspect/edit access to the semantic model through the Microsoft Power BI Modeling MCP. PBIP files are treated as version-control snapshots. Agents don't patch report JSON by hand.
Does it work outside Windows?
The CLI and several skills run cross-platform. Workflows that talk to Power BI Desktop (especially /bi-connect) are Windows-first because Power BI Desktop is Windows-first.
Do I need Claude Code?
No. Codex is the primary end-to-end validation baseline. Claude Code, GitHub Copilot, and Kilo Code are supported through compatibility adapters. Use whichever agent you already use.
Can I install only the MCP servers?
Yes. You can wire up Microsoft Learn MCP and Power BI Modeling MCP manually. bi-superpowers ships them together with the operational skills so the agent knows when and how to use them.
How do I update?
On the Claude Code marketplace: /plugin update bi-superpowers. On the npm/CLI install: super upgrade && super install --all --yes. Project-local plugins refresh with super recharge.
Is my data sent anywhere?
No. Everything is local: local files, local agent config, local MCP servers. There is no bi-superpowers SaaS backend.