AI assistants like Claude and ChatGPT are good at reading a situation and taking sensible next steps. What they have lacked is a safe way to see and act on the systems a warehouse actually runs on. The Model Context Protocol (MCP) changes that, and it’s why Dock Tempo includes an MCP server.
What MCP is, in one paragraph
MCP is an open standard that lets an AI assistant connect to a business system and use a defined set of tools: “list today’s bookings”, “find free slots”, “reschedule this booking”. The system decides what tools exist and who may use them; the assistant decides when to use them to answer your request.
What that looks like on a dock
At 7 am an ops manager types: “What needs me today?” The assistant calls yard_overview and answers: thirty bookings, twenty-eight already confirmed by Autopilot, one request for 05:30 before opening, one carrier who missed their 08:00. “Move the missed one to the first free slot after lunch and tell them.” The assistant calls find_free_slots, then reschedule_booking, and the carrier gets the new time by email.
Keeping it safe
- Same rules as people. Every tool goes through the same checks as the admin screens, so the assistant can’t double-book a dock or skip your rules.
- Read-only keys. Give reporting tools and curious colleagues a key that can look but never change anything.
- Everything logged. Each AI action is recorded with the key name and the reason it gave, next to Autopilot’s and your team’s decisions.
- Revocable. Revoke a key and that connection stops immediately.
Rules first, AI for the edge cases
Most bookings don’t need intelligence; they need consistent rules applied every time. That’s Autopilot’s job. AI earns its place on the exceptions: the short-notice request from a reliable supplier, the missed slot that needs a sensible new time. Dock Tempo’s AI booking review looks at exactly those, and you choose whether it recommends or acts.
The MCP connection is part of Dock Tempo Enterprise. Setup takes one line in Claude Code, or a short JSON snippet in Claude Desktop, Cursor or VS Code. See the docs.
