This MCP quickstart walks through the three steps every AI client needs to connect to the PlantDemand MCP server: initialize, tools/list, and tools/call. You can follow along with curl from any terminal, then move on to a specific client setup guide when you are ready to wire it into Claude, Cursor, Copilot Studio, or another assistant.
Before you start
- An active PlantDemand account with at least one plant configured
- A Server-Api-Key generated for your account (see SOP 4.1 in your customer documentation)
- A terminal with curl, or any MCP-capable AI client
Step 1: Initialize the session
Every MCP session starts with an initialize request. This confirms the server speaks MCP and reports what protocol features it supports.
curl -X POST https://plantdemand.com/mcp \
-H 'Content-Type: application/json' \
-H 'Server-Api-Key: <your key value>' \
--data '{"jsonrpc":"2.0","id":1,"method":"initialize"}'If you receive a JSON response with no error, your authentication is working and the server is reachable.
Step 2: List available tools
The agent then asks what tools are available. The server returns a list of tool names and input schemas.
curl -X POST https://plantdemand.com/mcp \
-H 'Content-Type: application/json' \
-H 'Server-Api-Key: <your key value>' \
--data '{"jsonrpc":"2.0","id":2,"method":"tools/list"}'The response includes the GraphQL-backed scheduling tool (order_dates) and the REST-backed operational tools for plants, materials, orders, customers, and field metadata. See the full tool reference for input schemas and example responses.
Step 3: Call a tool
The most common first call is order_dates, which returns scheduled order data for a plant. Replace plant_id with one of your own plant IDs.
curl -X POST https://plantdemand.com/mcp \
-H 'Content-Type: application/json' \
-H 'Server-Api-Key: <your key value>' \
--data '{
"jsonrpc":"2.0",
"id":3,
"method":"tools/call",
"params":{
"name":"order_dates",
"arguments":{
"plant_id":14,
"date_from":"2026-02-01"
}
}
}'The response contains scheduling information including quantity, delivery date, load rate, customer, material, and any extra fields configured for your plant. Permissions are enforced at the backend, so you only ever see data your account is allowed to access.
Step 4: Wire it into an AI client
Once the curl examples work, the next step is connecting an AI assistant. Pick the client you use:
- Anthropic Claude Desktop
- Claude Code (CLI)
- Cursor
- Replit Agent
- Microsoft Copilot Studio
- Azure AI Foundry
- Kiro
- ChatGPT
Common quickstart issues
401 Unauthorized
The Server-Api-Key header is missing or invalid. Confirm the header name is exactly Server-Api-Key (not Authorization) and the key value matches what was generated for your account.
403 Forbidden
Authentication succeeded but the API key does not have access to the requested plant. Plant-level permissions still apply through the MCP server. Check which plants the API key is associated with.
Empty tools list
If tools/list returns no tools, the server may not have completed initialization. Make sure you call initialize first in the same session.
Recommended agent instructions
When you connect an agent, give it a short system prompt similar to this:
Use the PlantDemand MCP server at https://plantdemand.com/mcp.
Send this header on every request: Server-Api-Key: <your key value>.
Begin each session by calling initialize.
After initialize, call tools/list and use only tools returned by the server.
For scheduling and dispatch lookups, use the order_dates tool.
For plant, order, customer, material, and field lookups, use the relevant REST-backed tools.
Do not invent values that should come from PlantDemand tools.
If a tool returns an error, report the server error clearly instead of guessing.Once the quickstart works end-to-end, see real-world use cases for prompt examples plant managers use day-to-day, or read about how PlantDemand fits into asphalt plant scheduling.
This guide is part of PlantDemand’s asphalt scheduling software hub for asphalt plant operations, scheduling, and sales management.