--- title: "Use — MCP Server for Claude, ChatGPT, Copilot & Custom Supply Chain Agents | bluefabric" description: "Connect Claude, ChatGPT, Copilot, Gemini, blueclip, or any custom agent to bluefabric over MCP. Give AI agents governed access to a clean, self-describing supply chain model spanning WMS, TMS, and ERP data." url: https://bluefabric.ai/use/ source: content/use/index.html --- [bluefabric](/)/ [Product](/#how)/ Use · MCP for any agent Pillars [Ingest](/ingest/) [Enrich](/enrich/) [Unify](/unify/) [Context](/context/) [Calculate](/calculate/) [Use](/use/) Layer 6 of 6 · Use # Give any AI agent _a supply chain brain._ Once your data is ingested, enriched, and mapped into the bluefabric model, it is ready to use. bluefabric exposes **clean supply chain context, trusted calculations, governed actions, and AI-readable object manuals** through MCP — so blueclip, Claude, ChatGPT, Copilot, Gemini, and custom agents can understand your supply chain without guessing from raw tables. Plug in once. Give every agent the same operational brain. [See bluefabric Live →](/demo/) 15-min walkthrough // one protocol · any agent Stop rebuilding supply chain context for every agent ## Every new agent _creates another integration problem._ One **connector** for ChatGPT. One **workflow** for Claude. One **schema** for Copilot. Another **custom bridge** for an internal agent. **That does not scale.** bluefabric gives every agent the same governed interface into your supply chain data layer: one model, one tool layer, one source of operational context. Models will change. Your supply chain brain should not. C ChatGPT→ custom connectorWMS · TMS · ERP A Claude→ different workflowWMS · TMS · ERP M Copilot→ another schemaWMS · TMS · ERP + Custom agent→ custom bridgeWMS · TMS · ERP 4 agents · 4 integrations · 4 failure modes One MCP layer. Every agent. ## Agents call bluefabric. _Not your systems._ bluefabric is exposed through the Model Context Protocol, so agents connect to one trusted supply chain interface instead of learning every source system directly. Agent request path Agent → bluefabric MCP → self-describing supply chain model → trusted calculations → governed actions → audit trail Agents do not need direct access to WMS, TMS, ERP, spreadsheets, portals, or legacy systems. **They call bluefabric.** One protocol. One schema. One audit trail. ![](https://pub-6d0b5b97762c4335b5b515672d21523f.r2.dev/img/technology/abstract-purple-blue-particle-wave-dark.jpg) Every object ships with its own AI manual ## The model doesn't just describe the data. _It tells the agent how to use it._ Every supply chain entity in bluefabric arrives with a structured manual the agent can read on first contact — what the object is, how to identify it, which states it can be in, what other objects it relates to, which tools to call, what filters to pass, and when to act. **No guessing. No prompt engineering against raw tables.** The model explains itself. Meaning What the object represents. Real supply chain semantics, not a column name. An _order_ means demand committed by a customer with an SLA, not "row 4821 in table X". Relationships How it connects. Orders, inventory, carriers, warehouses, customers, costs, exceptions, and service risk — pre-modeled, agent-readable, traversable. Tool guidance Which trusted call, when. Which deterministic calculation or action to invoke, which filters to pass, and when to escalate. The agent doesn't have to guess. Agent manual · Object Shipment v1 01 What it is A **physical movement of goods** from an origin to a destination, tied to one or more orders and executed by a carrier on a lane. Created when picked inventory is committed to leave a warehouse; ends when the goods arrive at the customer's dock. **Shipment is where most SLA risk materialises** — it connects warehouse activity, transport execution, and customer commitments. 02 Identity & key attributes idshipment\_id aliascarrier\_tracking\_no originwarehouse destinationcustomer planned\_etats actual\_etats statusenum (see lifecycle) weight · unitslb · count 03 Lifecycle planned→ picked→ loaded→ in\_transit→ out\_for\_delivery→ delivered | exception 04 Related objects Order Line Item Carrier Lane Customer Inventory Warehouse SLA Risk 05 Tools agents can call getShipmentRisk() getETA() getCarrierPerformance() detectDelayRootCause() 06 Filters date carrier lane customer warehouse status risk score 07 When the agent should act A shipment is _at risk_ when **planned\_eta + carrier history** suggests arrival after the customer commitment date. When in doubt, **call _getShipmentRisk_** before quoting an ETA. Never infer status from tracking strings alone — always resolve through the model. **The model does not just store data. It explains the supply chain to the agent.** shipment.sql shipment\_idVARCHAR statusVARCHAR etaTIMESTAMP carrierVARCHAR warehouseVARCHAR customerVARCHAR ← what does any of this mean to an agent? Raw tables ≠ supply chain knowledge ## Raw fields tell the agent what exists. _The manual tells it how to reason._ A raw shipment table exposes **shipment\_id, status, eta, carrier, warehouse, customer**. That is not enough. The agent still has to guess what the status means, which ETA matters, whether the carrier is reliable, which order is affected, whether the customer commitment is at risk, and which tool should be called next. **bluefabric removes that guesswork.** Every object arrives with its meaning, its relationships, its lifecycle, its tools, and its escalation rules — written for an agent to consume. Raw fields tell the agent what exists. The manual tells it how to reason. Built for blueclip. Open to every agent. ## Use bluefabric _wherever your agents already work._ bluefabric was designed to power blueclip's supply chain intelligence layer. But it is **not locked to blueclip**. Any MCP-compatible agent can use the same clean data, object manuals, trusted calculations, and governed tools. blueclip blueclip native Anthropic Claude native MCP OpenAI ChatGPT function calling Microsoft Copilot Azure AI · MCP Google Gemini tool use · MCP Perplexity Perplexity structured retrieval In-house Custom agents MCP · REST Pipelines Workflow automations API · webhook Agents do not need to learn every source system. They plug in through MCP and **immediately get clean, AI-ready supply chain context.** From generic chat to real supply chain work ## Chat is not the product. _Operational execution is._ Once connected through MCP, agents can move beyond generic Q&A. Ask Query clean operational context. Orders, SKUs, shipments, inventory, suppliers, carriers, costs, and risk — all connected through the model. Calculate Call trusted calculations. Fill rate, OTIF, landed cost, lead time variance, service exposure, exception impact — deterministic, traceable, repeatable. Act Trigger governed workflows. Route approvals, update systems, escalate risk, and log every action — through governed write-back, never raw access. Chat is not the product. **Operational execution is.** Planning agent queries **clean inventory & demand** Transport agent calculates **shipment risk** Customer service agent explains **order exposure** Warehouse agent understands **which exception actually matters** Your existing workflows get smarter ## Better context in. _Better workflows out._ bluefabric **does not force you to throw away the agentic workflows you already built.** It gives them better context. A planning agent can query clean inventory and demand. A transport agent can calculate shipment risk. A customer service agent can explain order exposure. A warehouse agent can understand which SKU, order, shipment, or exception actually matters. The workflow stays familiar. **The supply chain brain gets upgraded.** Better context in. Better workflows out. No direct system access. No uncontrolled agents. ## The agent gets power. _Your systems stay protected._ Agents do not need to touch your source systems directly. They connect through bluefabric, where every object, calculation, and action can be permissioned, governed, and audited. Governed agent request Agent request → bluefabric MCP → permission check → object manual → trusted tool → governed response → optional safe action **No direct WMS access. No mystery writes. No agent free-for-all.** [Explore Governed Write-Back →](/architecture/#governed-write-back) From clean model to live agent workflow ## Not a data dump. _A supply chain brain with instructions._ bluefabric turns fragmented supply chain data into something agents can actually use. 01 Ingest sources in → 02 Enrich clean · backfill → 03 Unify common model → 04 Calculate trusted methods → 05 · here Use MCP · any agent Not a data dump. A supply chain brain with instructions. ![](https://pub-6d0b5b97762c4335b5b515672d21523f.r2.dev/img/technology/abstract-magenta-particle-explosion-dark.jpg) Give your agents the manual ## Every object explained. _Every action governed._ Your agents do not need more raw tables. They need a clean model, trusted calculations, object-level guidance, and safe tools they can call. **bluefabric gives them all of it through one MCP layer.** Every object explained. Every tool described. Every action governed. [See bluefabric Live →](/demo/) [MCP architecture deep-dive](/architecture/#mcp-layer) [ ← Back: Layer 5 Calculate · Trusted Calculations How bluefabric gives agents deterministic, traceable KPIs instead of LLM guesses. ](/calculate/)[ ↑ Back to home The full bluefabric story Ingest, Enrich, Unify, Context, Calculate, Use — the six-layer supply chain brain for AI agents. ](/)