Production Plan Variance MCP Server
This server does not currently have a deployable release on Glama. Follow the repository's setup instructions to install and configure it for your MCP client, including any required credentials and local dependencies.
Production Plan Variance MCP Server
A small MCP server that exposes the Production Plan Variance & Issue Tracking
dashboard (erp.peopledesk.io/production-management/mes/ProductionPlanVariance)
as five read-only tools, backed directly by the DWH database:
Tool | Mirrors |
| The four tiles: Plan Lines / On Target / Variance Flagged / Issues Missing |
| The row grid (plan code, item, plant, qty, difference, reason, status) |
| The "Issue for Difference" reason config + escalation emails, per plant |
| Plant filter dropdown |
| Shop Floor filter dropdown |
It queries these DWH tables (verified against the live schema):
tblProductionPlanVarianceIssueArc, tblProductionPlanVarianceReasonArc,
tblPlantArc, tblShopFloorArc.
Why this shape
You asked for something anyone can use via a JSON file, without anyone entering database credentials. So the design is:
One server, hosted by you, holding the DB credentials in its own environment variables (
.env, never committed, never sent to clients).Everyone else just drops
mcp-config.jsoninto their MCP client (Claude Desktop, Claude Code, etc.) — it only contains a URL, nothing secret. The server is the only thing that talks to SQL Server.All queries are fixed, parameterized SELECTs defined in
src/queries.js— there is no free-form SQL tool exposed, so nobody using this MCP server can query outside the dashboard's own data or run writes.
Related MCP server: mcp-server
⚠️ Before you deploy: confirm DB permissions
While building this, I verified the schema (tables/columns) using your
mssql-test-server connection, but every actual row-level SELECT against
tblProductionPlanVarianceIssueArc / tblPlantArc etc. failed — schema
introspection worked, data queries didn't. That's almost certainly a
permissions gap on that specific SQL login (mcp_user), not a problem
with these queries. Before going live:
Create (or reuse) a SQL Server login with
SELECT-only rights on the DWH tables listed above.Test manually, e.g.:
SELECT TOP 5 * FROM dbo.tblProductionPlanVarianceIssueArc;Only then point
DB_USER/DB_PASSWORDin.envat that login.
Deploy
Option A — your own server (nginx/Caddy in front)
npm install
cp .env.example .env # fill in DB_SERVER / DB_USER / DB_PASSWORD
npm start # listens on :3300, endpoint POST /mcpPut it behind HTTPS on a domain reachable by your team, e.g.
https://mcp.yourcompany.com/mcp.
Option B — Vercel (already set up in this repo)
The api/mcp.js + vercel.json files convert the same server into a
Vercel serverless function, exposed at /mcp.
Push this folder to a GitHub repo (or run from the folder directly).
Install the CLI once:
npm i -g vercelFrom inside
production-variance-mcp/, log into your own Vercel account and deploy:vercel login vercel --prodIn the Vercel dashboard → your project → Settings → Environment Variables, add
DB_SERVER,DB_PORT,DB_DATABASE,DB_USER,DB_PASSWORD,DB_ENCRYPT,DB_TRUST_CERT(same values as.env.example), then redeploy so the function picks them up.Your MCP endpoint is now
https://<your-project>.vercel.app/mcp. Put that URL intomcp-config.jsonand hand it out.
Note: SQL Server must be reachable from Vercel's network (i.e. not
firewalled to only your office IP) — check with whoever manages
203.202.241.211 if the connection times out after deploying. Also add
some form of auth in front of /mcp (see note below) before this URL is
public — right now anyone with the link can call these read-only tools.
Distribute
Edit mcp-config.json, replacing YOUR-HOSTED-DOMAIN with your real
domain, then share that one file. Anyone adds it to their MCP client and
immediately has read-only chat access to this dashboard's data — no
install, no credentials, no SQL knowledge required.
Notes / things to double check on your end
The dashboard's "Machine" column (e.g. "Aromatic Unit Line - 01") wasn't found on
tblProductionPlanVarianceIssueArcin the schema I pulled — it's likely resolved via a production line/work-center table in your environment.list_plan_variance's query doesn't include it yet; add the appropriate JOIN insrc/queries.jsonce you confirm the source table (candidates:tblProductionLineArc,tblWorkCenterArc).get_variance_summary's "Issues Missing" logic assumes it means "flagged variance with no reason/status recorded yet" — confirm this against the app's actual business logic and adjustQUERY_SUMMARYif different.Add auth (API key / OAuth) in front of
/mcpbefore exposing it beyond your internal network — this build has none by default.
Tool Schema Changelog
Recent tool additions, removals, and schema changes observed during successful MCP inspections.
Loading tool schema changelog…
This server cannot be deployed
Maintenance
Related MCP Connectors
Read-only supply-chain decision support: forecasting, reorder policies, ABC-XYZ, data quality.
Securely query and analyze business data, dashboards, projections, alerts, and knowledge.
Live KPIs and shop data for Tekmetric and Shop-Ware. 29 read-only tools.
Manufacturing BI from your AI client: dashboards, OEE, spreadsheets, scheduling, text-to-CAD.
Related MCP Servers
- AlicenseAqualityDmaintenanceProvides read-only SQL Server access via MCP tools and a chat interface for natural-language database queries.4ISC
- FlicenseNot gradedqualityDmaintenanceProvides read-only operational tools (sales summaries, orders, top customers) over a multi-tenant dataset, with tenant isolation enforced server-side so no tenant parameter appears in the tool schema.-
- FlicenseNot gradedqualityCmaintenanceProvides remote read-only access to a Data Warehouse SQL Server, enabling users to query manufacturing OEE data, SBU lists, and run ad-hoc SELECT queries.-
- AlicenseNot gradedqualityCmaintenanceRead-only MCP server exposing Flowsery analytics and AI-detected issues via 17 GET tools, enabling natural language queries in compatible clients.MIT