Warehouse Management System MCP Server | TBO4 |

Most warehouse management systems give you dashboards. A dashboard answers the questions someone built it to answer, and every question after that is a change request with a lead time.

TBO4 ships with a native MCP Server: an open, standards-based interface that lets AI assistants query your live warehouse directly. Ask a question in plain English, get an answer from real-time data, not from a report someone has to build first.

What an MCP Server actually is

MCP stands for Model Context Protocol. It is an open standard, not a RIC Group invention, and that is the point.

A traditional API assumes the caller already knows what it wants. A developer reads the documentation, picks an endpoint, writes the code. All the intelligence lives in the client, and adding a capability means another development cycle.

MCP inverts that. The TBO4 MCP Server publishes a machine-readable description of every warehouse operation it exposes: what it does, what it needs, what it returns. An AI assistant discovers all of it on connection and works out which operations to call, and in what order, based on what you asked.

• Nothing is hard-coded on either side. That is why the questions you can ask are not limited to the ones we anticipated.
• 
  

What is the TBO4 MCP Server?

It is an interface that lets AI assistants read your live TBO4 warehouse data through the Model Context Protocol, an open standard for how AI tools discover and call operations in other systems. It runs alongside TBO4.

Works with the AI you already use

The TBO4 MCP Server is not tied to one AI vendor. Because the protocol is open, your choice of assistant is a configuration setting rather than a rebuild.

Anthropic Claude. Native MCP support. Connect Claude.ai or Claude Desktop with a URL and it discovers the full TBO4 tool set immediately. Best answer quality where direct connectivity is permitted.

ChatGPT and OpenAI. Connect through the same MCP interface.

Azure OpenAI. Recommended for the embedded TBO4 AI Assistant. Runs inside your own Azure resource, in the region you choose, under the Microsoft agreements you already hold.

Ollama. Fully on-premise inference for operations where no data may leave the network.

• Switch providers later and the warehouse side does not change.

Which AI assistants can connect to it?



Anthropic Claude, ChatGPT and OpenAI, Azure OpenAI, and Ollama for fully on-premise deployments. Because MCP is an open standard, any MCP-compatible client can connect. You are not locked into one AI vendor.

Inside TBO4, for the floor

The TBO4 AI Assistant is an embedded chat panel available on every TBO4 screen, switched on per user. Warehouse staff ask questions without leaving the system they are already working in.

"How many open picks today?" "What is the consignment for this SSCC?" "How do I do tote picking?"

• That last one matters. The MCP Server indexes your own process documentation, so "how do I" questions are answered from your documented procedure, not from a language model's general knowledge.
• Of the ERP transactions that failed to post back this month, how many relate to the customer with the highest order concentration? This chains ERP error listing, top customers, and pick detail.

Outside TBO4, for analysis

Operations managers, analysts and IT leads connect Claude or ChatGPT directly to the MCP Server and interrogate the warehouse from there. No widget, no code, no integration project. A connector URL and permissions.

This is where the open-ended nature earns its keep. Questions that chain several parts of the operation together, which nobody would have commissioned a dashboard for:

Which bins see the most picking activity for the SKUs that also show the highest cycle count variance?

Of the ERP transactions that failed to post this month, how many belong to our largest customer by order volume?

Which carrier is costing us most per consignment, and has that changed since last quarter?

Do we need a custom integration project?



No. The MCP Server is native to TBO4. For an external assistant such as Claude, connection is a URL and a permission set. There is the embedded TBO4 AI Assistant as a configuration option.

The Coverage

The MCP Server exposes your warehouse as a set of named, typed operations across every operational domain in TBO4:

Inventory and bins - Stock on hand per bin, bin contents, bin attributes, zones, default pick and putaway locations

Items, barcodes, serials - Item master, all barcodes, batch and serial numbers with expiry, unit of measure conversions, kits

Picking - Live pick queue by warehouse or across all sites, full pick detail with required and picked quantities per bin, operator productivity

Receiving and putaway- Receiving queue, purchase order detail, goods received by date range, putaway queue and status

Packing and freight - Pack queue, SSCC carton contents, consignment detail with carrier and addresses, freight spend by carrier, ASNs

Transfers and replenishment - Open transfer requests, replenishment against reorder point, transfer history

Stocktake and cycle count - Count lines, section progress, variances above any threshold you set

Production and jobs - Production orders, warehouse job headers and lines

RFID Tag records and activity, via the SATO integration - Audit trailsPick, pack, transfer, receiving, putaway and adjustment history by user and date

Analytics - Trends and top-N across picking, receiving and inventory, by day, week, month or between any two dates

Documentation - Your own TBO4 process guides, searchable by topic and SOP's

How do we control who can ask what?

Access is controlled at two points. Who can open the assistant inside TBO4 is set per user in TBO4 user profile settings. Which operations an external assistant may call is set per operation in the AI client, so you choose what is always allowed, what asks each time, and what is blocked.

Can we add our own questions and reports?



Yes. The analytical query catalogue, the fast-path shortcuts and the process documentation are all configuration files. Add to any of them without a software release. Your team can do it, or we can.