From idea to app: designing a Fishbowl tool with ILC MCP and CloudPages
Some Fishbowl problems need a tool someone on your team uses every day: a packing list printer, a receiving check-in, a label station. We build those as CloudPages, our tools that run inside Fishbowl. With ILC MCP connected to your Fishbowl, the first picture of that tool can come out of the same conversation where the idea comes up, using your own data.
Getting momentum
Most ideas for a better tool never get built. Someone on the team knows exactly what they want, but it lives in their head or in an email, and nobody approves a project they can’t picture. A rough draft with your own data on it changes that. The warehouse lead points at what’s missing, the owner sees the time it would save, and the idea turns into a project.
Designing it with your data
With ILC MCP connected, you describe the problem in Claude or ChatGPT, and the agent looks at your actual Fishbowl data while you talk. It reads your data and can’t change it.
Say your shipping team tells you, “I need a packing list printer.” Fishbowl’s standard packing list doesn’t show what your customers ask for.
- The agent finds where the data lives. The shipment, its cartons, the order lines, the customer’s PO number, and the lot and expiry for each item.
- It checks your real cases. It notices that some shipments go out in several cartons, that a few parts are lot-tracked, and that one large customer wants their own part numbers on the list. Those are the cases that usually appear after the build starts.
- It drafts the packing list. A layout filled in from one of your real shipments, and a simple diagram of where each field comes from. In Claude, that can be a printable mockup in the chat.
- It saves the queries. The queries behind the packing list become saved queries your admin can review and run, so the data side is tested before anyone writes the tool.
You go back and forth until the packing list looks the way your team and your customers need it.
From draft to CloudPage
The draft is a specification, and our developers take it from there. They build the CloudPage inside Fishbowl, with your team’s logins and permissions, handle the edge cases the agent found, test it against your data, and maintain it after it goes live.
The agent’s mockup is a draft. Permissions, printers, error cases, and two people printing at once are the developers’ job. What the draft gives you is something your team can react to, and a clear starting point for us, with the edge cases already found.
When this is the right fit
This works best for a job that:
- Someone does many times a day, where a chat would be too slow.
- Needs a person to decide something, so an unattended integration doesn’t fit.
- Uses data that’s already in Fishbowl, shown or combined in a way Fishbowl’s own screens don’t.
If it runs on a schedule with nobody watching, it’s an integration. If it comes up once, a chat is enough. We cover that split in when not to use an MCP.
If there’s a tool your team wishes Fishbowl had, tell us what it would do.
Related: