SkyPlanner MCP server
MCP server for production scheduling
Connect Claude, ChatGPT or Copilot to SkyPlanner through the Model Context Protocol. Your AI agent reads orders, capacity and the finite-capacity schedule, changes them and asks Arcturus AI for a new schedule. Included in the free 30-day trial, so you can prove it with your own data before you buy.
Your sheet is now a live schedule in SkyPlanner: 4 workstations, 12 products and 38 open orders.
The sheet had no shift times, so I used 7 to 15 and 15 to 23 on weekdays. Tell me if that is wrong.
Three orders are planned to finish after their due date. Want to see which ones?
If the order comes in today, it is planned to finish on Thursday next week.
No current order becomes late because of it. This was a trial run in the sandbox, so your live schedule is unchanged.
Want me to check what a delivery by Tuesday would push back?
Laser cutting 2 is your bottleneck. It has the longest queue, and most of the late orders are waiting for it.
| Workstation | Jobs waiting | Queue ends in |
|---|---|---|
| Laser cutting 2 | 14 | 9 working days |
| Bending 1 | 5 | 3 working days |
| Powder coating | 3 | 2 working days |
Laser cutting 1 can run six of those jobs. Should I try moving them in the sandbox?
Three orders are planned to finish after their due date:
| Order | Customer | Due | Late by |
|---|---|---|---|
| 24-1187 | Atlas Metalworks | Wednesday | 2 days |
| 24-1192 | Summit Frames | Thursday | 1 day |
| 24-1201 | Lakeside Pumps | Friday | 4 hours |
All three are waiting for Laser cutting 2, which is overbooked on Thursday. Moving order 24-1210 to Friday would free 6 hours.
Order 24-1215 is in SkyPlanner with the default routing of the product: laser cutting, bending and powder coating.
Laser cutting starts on Monday, and the order is planned to finish on Wednesday, two days before delivery.
On-time delivery is now the top scheduling goal. Live is unchanged.
| Live | Sandbox | |
|---|---|---|
| Late jobs | 7 | 2 |
| Workstations with late jobs | 3 | 1 |
Should I move the rule to live?
- Claude, ChatGPT and Copilot
- Read or read-and-write access
- Arcturus AI computes every schedule
- Included in the 30-day trial
Proof of concept
Prove it with your own data, during the trial
No project, no consultant, no developer. The Getting Started wizard offers the MCP connection on the first day of your trial. Connect the AI assistant you already use and let it do the whole setup with you.
Hand it your data.
Connect your assistant in the Getting Started wizard. Export orders, routings and workstations from your ERP, even as an incomplete spreadsheet. Your assistant models them in SkyPlanner and asks about what is missing.
Let Arcturus schedule it.
Ask for a schedule. Arcturus AI builds a finite-capacity plan from your real orders and capacity. A machine shop’s unfamiliar data became a scheduled production model in a few hours.
Ask what you need to know today.
Which orders are late? Where is the bottleneck? What if we add a shift? Answers from your own production, before you decide anything.
Read and write
Your assistant reads the plan. Arcturus AI still computes it.
Your assistant sees everything a planner sees and changes orders, shifts and settings. The schedule itself is always built by Arcturus AI under your rules and priorities.
Your assistant reads
Orders · Load · Schedule
- Orders, production steps and late work
- Workstation queues and daily load
- Shifts, downtime and time logs
- Products, routings, stock and purchases
- Scheduling rules and SkyPlanner help
SkyPlanner APS
Arcturus AI schedules
- Finite capacity
- Priorities
- Setups
- Materials
- Shifts
- Sandbox
Your assistant asks for a reschedule. Arcturus computes it in seconds, exactly as it does for your planners.
Your assistant writes
Data · Schedule · Settings
- Create and update orders, workstations, shifts, downtime, customers, purchases and stock
- Schedule, reschedule and move jobs
- Change scheduling rules and settings
What to ask
Ask it the way you would ask a colleague
Every role on the shop floor and in the office gets its own answers from the same live production data.
Production planner
Which machines have more work than shift time next week, and which orders slip because of it?
It reads the live schedule, names the bottleneck and moves the jobs or reschedules.
Sales
A customer asks for a quote. If they order today, when can we deliver?
It places the order in a sandbox copy of production, schedules it and shows where it lands and what it pushes. The date in your offer comes from real capacity.
Production manager
Would a Saturday shift on Welding 1 be enough next week? Try it in the sandbox.
It adds the shift to the sandbox, reschedules and tells you which orders come in on time and which still slip.
Foreman
Lathe 2 is down. Log a breakdown until noon tomorrow and reschedule.
It records the downtime and asks Arcturus AI to rebuild the plan around it. The shop floor sees the new order of work at once.
Getting started
Here is an export from our ERP. Model our workstations, routings and orders.
Hand it an incomplete spreadsheet. It builds the production model in SkyPlanner, asks about what is missing and schedules it.
Every morning, on a schedule
Every weekday at 6:30: which orders are late today, which machines are overloaded this week and what changed since yesterday.
Set it up once as a scheduled task in Claude. The briefing is ready before the morning meeting, from the live schedule, without anyone logging in.
Sandbox
Simulate in a copy of production. Move only what worked.
Every SkyPlanner cloud installation has a sandbox, a copy of your real production. Your assistant runs scenarios and tunes Arcturus AI there, with production untouched, and then carries the settings you choose into production.
Run the scenario
A rush order at top priority, a Saturday shift on the bottleneck, a new machine, a quote that may become an order. Your assistant refreshes the sandbox from production, makes the change and asks Arcturus to reschedule.
Tune Arcturus AI
Change scheduling direction, priorities, locked jobs or how a job splits across machines, and see what each rule does to delivery dates before anyone on the shop floor notices. The assistant explains every setting from SkyPlanner’s own help.
Move what worked to production
Compare the result, which orders moved earlier and which later, and then carry the chosen settings and company-wide scheduling rules into production. The rest of the sandbox stays where it is.
Nothing to install, easy to integrate
Hosted by SkyPlanner, connected in minutes
The MCP server runs as part of SkyPlanner. Its connection page has ready instructions for Claude, ChatGPT, Claude Code and Codex, and any other MCP-capable application connects with the same server address. You sign in with your SkyPlanner account and decide per user whether the assistant reads or also writes, for production and for the sandbox separately.
Works with your ERP
SkyPlanner integrates with any ERP or MES system through its REST API. An assistant that also has your ERP connected works across both: it can check an order in the ERP and see where it sits in the SkyPlanner schedule in the same conversation.
Integrators build faster
An integrator rarely knows the production use cases behind the data. With Claude Code or Codex connected to SkyPlanner, the agent writing your REST API integration sees SkyPlanner the way a planner does: what each field means, which orders, routings and workstations landed and where, and what Arcturus AI did with them. It pushes test data into the sandbox and checks the result at every step, so the integration is verified while it is built.
- Business Central
- Dynamics 365 SCM
- NetSuite
- Odoo
- SAP Business One
- SAP S/4HANA
- Epicor Kinetic
- IFS Cloud
Setup guide
Connect your assistant step by step
Copy the server address and follow the steps for your assistant. The same address and steps are on the MCP page in SkyPlanner.
Server address
mcp.skyplanner.app/mcpBefore you start
- MCP connections are allowed on the MCP page in SkyPlanner. If the page says they are turned off, turn them on first.
- In a sandbox, connect in your live SkyPlanner. The same connection works with the sandbox too.
- If the MCP page says the server address is not set up for your installation, contact SkyPlanner support.
- In Claude, open Customize → Connectors, select + Add and then Add custom connector.
- Name it SkyPlanner, paste the server address and select Continue. Keep the suggested sign-in settings and select Add.
- Sign in to SkyPlanner and approve the access. In a chat, turn SkyPlanner on under + → Connectors.
Watch the video: connecting Claude
Claude Code
Run this command in your terminal:
Then type /mcp in Claude Code, choose skyplanner and sign in to SkyPlanner.
- In ChatGPT, open Plugins, select Add and then Add MCP server.
- Name it SkyPlanner, paste the server address and create the plugin.
- Install the plugin, sign in to SkyPlanner and approve the access. In a new chat, type @ and choose SkyPlanner.
Watch the video: connecting ChatGPT
Codex CLI
Run this command in your terminal:
Sign in to SkyPlanner when your browser opens. If it does not open, run:
- Open Copilot Studio and choose the agent to add SkyPlanner to, or create a new one. If you have no access to Copilot Studio, ask your Microsoft 365 administrator.
- On the agent’s Tools page, select Add a tool, New tool and then Model Context Protocol.
- Name it SkyPlanner, paste the server address, choose OAuth 2.0 and Dynamic discovery, and select Create.
- Create a connection, sign in to SkyPlanner and approve the access. The tools can take a while to appear. If they do not, wait a moment and refresh the page.
- So that everyone signs in with their own SkyPlanner account, set the tool’s credentials to End user credentials. Then publish the agent to Microsoft 365 and Microsoft Teams for others to use.
- Add a remote MCP server in your client and paste the server address.
- Sign in to SkyPlanner when asked and approve the access.
Requirements
The client must support remote MCP servers over HTTP with OAuth sign-in, and Client ID Metadata Documents (CIMD).
Want a demo or help connecting your assistant? Book a meeting
Questions
Frequently asked questions
Does this replace Arcturus AI or SkyPlanner’s own assistant?
No. Arcturus still computes every schedule, and SkyPlanner’s built-in assistant keeps working inside SkyPlanner. The MCP server brings SkyPlanner into the AI application you already use, next to your other tools and files, and lets it ask Arcturus to reschedule.
Do we need a developer or an IT project?
No. The MCP server is hosted by SkyPlanner. Open the MCP page in SkyPlanner, follow the ready instructions for your AI application and sign in with your SkyPlanner account. The Getting Started wizard offers the same connection at the start of the trial.
Is it safe to give an AI assistant access to our production data?
Yes. You decide per user whether the assistant reads or also writes, for production and for the sandbox separately. Every call is logged, and the sandbox lets you try changes in a copy of production first.
Is it included in the trial?
Yes. The MCP server is part of the free 30-day trial, so you can connect your assistant and test SkyPlanner with your own data from day one.
Start with your own data
Start a free 30-day trial, connect your assistant and ask it the first question about your production. The MCP server is included.