External AI
The same toolkit Alan uses inside TuringDock is exposed via the Model Context Protocol. Connect Cursor, Claude Desktop, your in-house assistant, or anything that speaks MCP, and your AI queries live inventory, BOMs, supplier status, customer orders, the schedule, and the books.
Cursor config · settings.json
{
"mcpServers": {
"turingdock": {
"url": "https://mcp.turingdock.com/mcp",
"headers": {
"x-workspace-slug": "elastic-energy"
}
}
}
}Then, inside Cursor
You: what should I count today on the Elastic Energy shop?
Cursor: Pulling your top count recommendations…
EM-100, SCREW-M3, VREG-3V3, HEATSHRINK-3, ATMEGA328P. PCB-RAW is also urgent (current 24 vs reorder 60); you counted it 3 hours ago though.
The MCP server exposes the same eleven tools that power Alan in Slack: inventory levels, open POs, material availability, supplier details, FX rates, schedule drafting, task state machine, schedule summaries, count recommendations, count logging, report summaries + anomalies, and demand forecasting.
Add one config block to your MCP client and every tool surfaces as a callable tool inside Cursor, Claude Desktop, or whatever you use.
When Cursor logs a cycle count via the MCP server, it writes the same inventory_count_event the web app or Slack bot would. When Claude Desktop marks a production task done, the stock movements post atomically against the BOM snapshot. The MCP surface isn't a read-only mirror; it's the same action surface, addressable from a different client.
Each MCP connection is bound to a single workspace via a slug in your config. Calls act on behalf of the workspace's primary admin during the beta. A per-user token model is coming next so the conversation in your AI client attributes correctly inside TuringDock's audit log.
TuringDock is in public beta. The premium tier (including this MCP surface) is being shaped with the first manufacturers using it. Book a 30-minute call with the founder to talk through what your team would actually use it for, and how we'd price it for your shop.
Legacy MRPs treat data as a destination. You go to them, you click through screens, you build reports. The system has no idea your AI tool exists.
TuringDock treats your manufacturing data as a context source for any AI you already trust. The same tools Alan uses live, work in your editor too. You ask Claude Desktop "what should I count today," it queries your live shop, you act on the answer without switching apps.
Build what's next. Leave the MRPing to our AI.
Leave the MRPing to our AI. Sign up free, set your shop up in an afternoon.
Get started for freeLast updated 2026-05-25