PlatformAI
AI that knows your plant and asks before it acts.
Courage Chat answers from your CRM, sales, quality and process-mining records and never changes anything without your confirmation. Agents outside CourageMFG reach issues and CAPAs over MCP, signed in as the person using them, with every change recorded.
Courage Chat
Ask the way you'd ask a colleague: what's closing this month, which items lost margin, what's open with this customer. Courage Chat looks it up through CourageMFG's own API, then answers with the records and numbers behind it, each one a link.
When an answer points to a change, such as a follow-up task, a new close date or a new issue, Courage Chat proposes it. You see what will change and confirm it, and the app makes the change itself, with your name on it.
- Runs as the signed-in person, through CourageMFG's own API
- Reads run straight away; writes are always proposed, never made by the model
- Works with Claude or Gemini, whichever your company chooses
- Today in CRM, marketing, sales analytics, quality and process mining; more areas to come
In marketing it reports how the last emails did and drafts the next one for you to review, test and send. Designing an email in Claude itself? Paste the HTML into CourageMFG: it's checked, then sent with your footer and unsubscribe link. Marketing and email
Which customers bought less this quarter than last, and is anything open with them?
sales_analyticsNet sales by customer, this quarter against last, from Business Central value entriessearch_recordsOpen opportunities, quotes and issues for those customerspropose_task Follow up before the quote expires, due FridayAn ontology, so answers are grounded
Language models are only as good as what they can look up. CourageMFG gives them a typed model of the plant: 47 object types, 78 link types and 19 action types, from customer and item to lot, machine, issue and CAPA. A question walks the links instead of guessing at joins.
MCP for Claude, Copilot and your own agents
CourageMFG publishes 9 tools over the Model Context Protocol for finding, reading, raising and moving issues and CAPAs. Connect Claude, Microsoft Copilot or an agent you build, sign in with Microsoft Entra, and it works as the person using it, with every change recorded under their name. More tools are coming as each area opens up.
No vendor lock-in for AI: the model is yours to choose and change, and so is the data it reads.
| Tool | What it does |
|---|---|
search | Find issues and CAPAs by number, title or a linked Business Central key |
get_record | Everything about one issue or CAPA, with its links |
get_related | What a record connects to across the ontology |
my_work | The person's own open issues and CAPAs |
pareto | What's driving issues, by category, work center, vendor, item or lot |
create_issue | Raise an issue linked to BC records |
triage_issue | Set severity and owner |
escalate_to_capa | Open a CAPA that inherits the issue's links |
transition_record | Move a record to its next allowed stage |
Guardrails that come from the design
Not a policy document: how the product is built.
Your permissions, never more
Every tool call goes through CourageMFG's API as the signed-in person, with the same rules and the same history as the app. There is no hidden super-user for the AI.
People confirm every change in Courage Chat
In Courage Chat the model can only propose a write. The person sees what will change and confirms it; the app then makes the change and records who made it.
Your model, your keys
Courage Chat runs on the model provider your company chooses, with your key, in your deployment. Switch providers without changing how anyone works.
Ask your own plant a question.
We'll connect CourageMFG to a copy of your Business Central and you can try Courage Chat on your own customers, items and orders.