weclapp MCP or Steward: safe AI access to weclapp for your whole team
Looking for a weclapp MCP server? Compare a DIY MCP setup with Steward: one shared connection, separate rights and an API token the AI never sees.
Why people search for a weclapp MCP
You run your business on weclapp. You use ChatGPT or Claude every day. The obvious question follows: can I just ask the AI about my orders, articles and invoices?
The technical answer people find is MCP, the Model Context Protocol. An MCP server is a small adapter that lets an AI assistant call another system’s API. At the time of writing, we know of no official weclapp MCP server. The options you will find are community projects on GitHub and paid builders that host a weclapp MCP for you.
Those projects prove that AI and weclapp work well together. They were mostly built for one person at one desk, though. A company has a sales team, a warehouse, a bookkeeper and a managing director, and each needs something different. That is the gap Steward fills.
How a typical weclapp MCP setup works
Most weclapp MCP servers follow the same recipe:
- Someone creates a weclapp API token in their personal weclapp settings.
- They paste that token into a configuration file or a hosted MCP dashboard.
- They connect the MCP server to their own AI app, such as Claude Desktop, ChatGPT or Cursor.
- The AI can now call the weclapp API with that token, which means with that user’s full rights.
For a developer exploring the API, this is perfectly reasonable. For a business, four questions come up quickly.
Four questions every business owner should ask
Whose rights is the AI using? A weclapp API token acts as the person who created it. If that person is an administrator, every colleague who uses the same setup effectively becomes an administrator as well.
Who sees the token? It ends up in configuration files, password-manager notes, chat messages (“can you send me the token again?”) and laptops. When someone leaves the company, do you know every place it went?
What happens before something changes? Many MCP servers expose write tools such as “create sales order” or “update article”. The AI calls them, and the change happens. There is often no step where a second person checks the exact change first.
Where is the history? Each person’s AI conversations live in their own account. When a price changes unexpectedly, there is no shared record of who asked, what the AI proposed and who approved it.
None of this is a flaw in MCP itself. It is simply what you get when a single-user tool is used by a whole team.
How Steward approaches weclapp differently
Steward is an AI workspace built for teams. It connects to weclapp directly. There is nothing to install and no server to run, and the controls a company expects are built in.
Connect once, share deliberately
A workspace owner or administrator connects weclapp once with the company address and an API token. The connection then belongs to the workspace, not to someone’s laptop.
Colleagues never receive the token. Instead, you decide which projects may use the connection and which people or teams may work with it. When someone leaves, you remove their access in one place. The token does not need to be hunted down.
Reading, changing and approving are separate rights
Project roles. Viewers follow results, Contributors use configured connections and prepare changes, and Managers approve. Invited project members need no separate connection grant.
Changes are proposals first
You need Manager access to the proposal’s project to approve or reject changes. Contributors can prepare changes and review the details. Approval rechecks the executor’s current project access and the enabled connection before execution.
For changes that should always get a second look, such as creating sales orders, approval rules on the connection keep the review in place even if someone switches a chat to Automatic.
The AI never sees the token
Steward stores the token encrypted and never shows it again, not even to the administrator who entered it. When the AI needs weclapp data, it asks Steward’s backend to make the request. Steward adds the credentials and checks the request against your permissions before it goes out. The token never appears in a conversation, a prompt or code the AI writes.
One shared history
Conversations live in shared projects. The question, the AI’s reasoning summary, the proposed change, who approved it and the result all stay in the same place. That makes it possible to answer “why did this price change?” months later.
weclapp MCP and Steward side by side
| Typical weclapp MCP server | Steward | |
|---|---|---|
| Setup | Install or host a server, edit configuration, connect your AI app | Enter address and token once in Settings |
| Who can use it | Whoever has the configuration | People and teams you grant access to |
| Whose weclapp rights apply | The token owner’s, for everyone | The token’s rights, narrowed per person by Steward permissions |
| Token visibility | In configuration files or the hosting dashboard | Encrypted, never displayed, never given to the AI |
| Writes | Usually execute immediately | Proposals with approval by default |
| “Always review” rules | Not typical | Per connection, also in Automatic mode |
| History | Each person’s own chat app | Shared project history with approvals |
| Recurring checks | Build it yourself | Scheduled tasks, e.g. daily at 17:00 |
What your team can ask on day one
Steward comes with weclapp reads already mapped, including articles, customers, sales orders and invoices. Some examples:
Which sales orders from last month are still unpaid? Group them by customer and show the overdue total in euros.
List the 20 articles with the lowest stock compared with their average monthly sales.
Compare this quarter’s revenue per customer with the same quarter last year and highlight anyone who dropped by more than 30 %.
Prepare price updates for these five articles. Show me the old and new prices before anything changes.
You need Manager access to the proposal’s project to approve or reject changes. Contributors can prepare changes and review the details. Approval rechecks the executor’s current project access and the enabled connection before execution.
You can also turn a useful question into a scheduled task. For example, a daily unpaid-orders report at 17:00 in Read only mode, so it can never change a record.
When a plain MCP server is still the right tool
If you are a developer experimenting alone with the weclapp API on a test system, a lightweight MCP server is a fine choice. Steward is built for the moment the experiment becomes daily work for several people, when it matters who may see what, who may change what and who said yes.
Get started
- Connect weclapp in Settings → Connections.
- Enable it for a project and grant colleagues the rights they need.
- Ask your first question in Read only mode, then try a reviewed change in Ask for approval.
Want to know why we never recommend pasting a token into a chat window? Read Should you paste your weclapp or Shopify API token into ChatGPT?