S-Curve Labs

S-Curve LabsMCP servers

Series MCP · 5 tools

Your assistant can write the formula. Now it can see the model.

These MCP servers let Claude, GitHub Copilot and other assistants work inside Power BI models, Power Query, SQL Server, Primavera P6 schedules and Power Apps.

Each one is made for real work files. Changes come with a dry run, a snapshot or a config switch first.

See them work before you install anything: try each server on a sample project, with the exact tool call and response for every question.

The servers

Five servers, each for one tool you already use. Install from PyPI and add it to any MCP client. MIT licensed.

  1. dax-quax

    Power BI

    What does every object in this model cost, what depends on it, and can it be deleted?

    Measures a tabular model's columns, tables and measures, traces every dependency, and sorts each object into remove, review or keep. The MCP mode is read-only, so an assistant can explain what is safe to delete and why.

    pip install "dax-quax[file,report,mcp]"
    PyPITry it

  2. letin

    Power Query

    Which queries depend on this one, and can I rename it without breaking the rest?

    Dependency graphs, reference-safe renames, bulk edits, lint and query folding checks for M in Excel workbooks and Power BI projects. It runs a query to show what each step really returns. Every write takes a snapshot first and shows a diff.

    uvx letin
    PyPITry it

  3. sqlglass

    SQL Server

    What does this query actually touch, and what does the optimizer think of it?

    Schema-aware query building, lint, read-only execution and plan summaries for SQL Server and Azure SQL, plus a .sql query library kept in git. An UPDATE or DELETE comes back as the SELECT that shows which rows it would change.

    uvx sqlglass
    PyPITry it

  4. longpath

    P6 & MS Project

    Where is the critical path, and did this update quietly move it?

    Reads P6 .xer, P6 XML and MS Project files the way a scheduler would: its own CPM engine, float, driving paths, the DCMA 14-point assessment, logic checks, update-to-update comparison and S-curve data. Schedule files are only ever read.

    uvx longpath
    PyPITry it

  5. fxray

    Power Platform

    What is inside this solution, and what changed since the last one?

    Wraps the Power Platform CLI to read canvas apps as text (screens, controls, Power Fx formulas, data sources), describe Dataverse tables, run read-only FetchXML and diff two solution exports. Pack, import and publish default to a dry run.

    uvx fxray
    PyPITry it

How they work

What "careful" means for a server an AI assistant can call.

Read first

Inspecting, tracing and checking work without changing anything. Most of each server is read-only.

Plan before writing

Writes return the exact change first: a diff, a preview SELECT, or the command a dry run would execute. Nothing is applied until you pass dry_run=false or turn writes on in the config.

Runs on your machine

Each server is a local process your assistant talks to over stdio. Your workbooks, models and schedules stay where they are.

Setup and troubleshooting