weclapp in Microsoft Copilot Studio: ERP agents for IT service providers

Does your company live in Microsoft 365? Then your ERP can work there too.

If Teams, Outlook and SharePoint structure your workday, the obvious question isn't "which AI tool should we buy?" but: why doesn't my ERP show up where my team already works?

That's exactly what wals.pro AI 4 weclapp does. The platform connects your weclapp tenant to the AI client of your choice via the Model Context Protocol, and Microsoft Copilot Studio is listed as a supported client on the connection page. Your self-built agent doesn't just answer questions about documents, it works live in the ERP: looking up sales orders, checking contracts, logging time, updating tickets.

The point is that your team keeps its environment and still gets full, traceable weclapp actions.

What an ERP agent in the Microsoft stack actually does

wals.pro AI 4 weclapp comes with 51 tools across 146 entities, 35 of which are write actions. For IT service providers these aren't abstract numbers, they're everyday work:

• Service and support: summarize a customer's open tickets, add comments, log effort to the right task.
• Contracts and billing: check recurring billing, look up contract items and terms before the invoice goes out.
• Sales: from lead to quote and sales order to invoice, including PDF.
• Reporting: aggregate key figures server-side instead of exporting lists and adding them up in Excel.

The appeal for Microsoft shops lies in the combination: the agent that already knows your documents and calendar now knows your order backlog too.

How the connection works technically

wals.pro AI 4 weclapp is a secured execution layer without its own language model. It maintains the connection to your tenant, enforces permissions and approvals, and writes every call to the audit log. The language model stays your contract, in this case Microsoft.

Setup on our side takes three steps and usually about half an hour:

1. Register and connect weclapp. Sign in at ai.wals.pro and enter your weclapp URL and API key.
2. Connect the AI client. Sign-in runs via OAuth 2.1, not via a copied static token.
3. Set permissions. Start read-only. You enable write permissions once your team knows the behavior.

On the Copilot Studio side, according to Microsoft's documentation, it works through your agent's tools menu: go to Tools, create a new connection of type Model Context Protocol, enter the server URL, choose OAuth as authentication and go through the sign-in once. The redirect URL that Copilot Studio shows you must be stored in the corresponding app registration. After that, per Microsoft's description, the connection is available as a tool in your agent. The sign-in flow is where things most often get stuck in practice: if you get stuck, write to us and we'll take a look together.

Sounds like an afternoon of work? Then rather book a short intro call. We're currently supporting the first Copilot Studio integrations and know the spots where the sign-in flow tends to get stuck.

Free choice of AI client, same rules for writing

Important for everyone working in a regulated environment: the security path doesn't depend on the AI client. Whether your team works through Copilot Studio, Claude or Codex, every write operation runs identically:

1. Read first. The agent reads the current state before changing anything.
2. Show a preview. Every change comes as a preview: what changes, what stays.
3. Wait for approval. No write without explicit OK. The approval is payload-bound via SHA-256 and applies to exactly this action with exactly these values.
4. Verify the result. After writing, the agent reads the record again and confirms.

Every call lands in the audit log, and there are no delete tools for your business data. "The AI suggests. You decide." is therefore not a marketing line but wired into the platform's approval logic. The agent in Copilot Studio gets no special rights just because it comes from Microsoft.

Your process knowledge stays with you

The strongest reason not to make the front end a matter of principle: the real asset isn't the tools but the knowledge of the order in which to use them. In wals.pro AI 4 weclapp, that knowledge lives in the skills, installable workflow packages for complete ERP processes. They follow the open Agent Skills standard and run in Claude, OpenAI Codex, VS Code and Microsoft Copilot Studio, among others.

In practice, this means: if your company switches AI providers in two years, your process knowledge moves with you. You've invested in a way of working, not in a subscription.

Why this is especially interesting for IT service providers

If you look after weclapp customers, this is twofold. First, a tool for your own operations, because tickets, time and contracts are exactly the objects your daily business is made of. Second, a topic your customers will raise anyway, especially the Microsoft-heavy ones among them.

Our experience from our own operations: the aha moment doesn't come from a feature comparison, but when someone types the first prompt themselves and the agent visibly works cleanly. That's exactly why starting with a small, read-only agent is a better idea than a big concept.

For weclapp partners who want to offer this under their own banner, we're currently building a Founding Partner program. It's a program with a limited number of spots, not an off-the-shelf product. If that interests you, talk to us directly.

Get started

👉 Start for free, open beta until 31.12.2026: ai.wals.pro

👉 Want to set up the Copilot Studio integration together or talk about a partnership? Book a call: wals.pro/termin