Using ILC MCP
Once connected, you work by describing what you want. ILC MCP reads Fishbowl, answers your questions, and — for anything that changes data — proposes the exact change and waits for your approval.
The pattern
Section titled “The pattern”Reads and writes follow the same shape, with one extra step for writes:
- Ask for what you want in plain language.
- Review what ILC MCP found, or the exact change it proposes.
- Approve — for a change, open the approval link, sign in, and run it. Reads need no approval.
Nothing is written until you approve. If a proposal is wrong, correct it in plain language and ILC MCP revises before anything runs.
Looking things up never changes anything and needs no approval.
- Asking questions — look up anything in Fishbowl.
- Reporting — build and open reports and dashboards.
Changes
Section titled “Changes”Anything that writes to Fishbowl is proposed and approved first. See Update Fishbowl:
- Placing orders — sales, purchase, transfer, manufacture.
- Managing inventory — receive, move, adjust, scrap, fulfil.
- Importing data — parts, customers, vendors, pricing, BOMs, and more.
Tips for good results
Section titled “Tips for good results”- Say the outcome, not the mechanism. “Order a month’s supply from our usual vendor” works better than naming an import type.
- Confirm in your own words. “Yep, that looks right — send it” is enough.
- One job per request keeps the review easy to read.
- Correct, don’t restart. If a number is off, tell ILC MCP; it adjusts the same proposal.
- Your model and its settings matter. The AI client, the model you pick, and its thinking or reasoning settings all affect how accurate and complete the answers are. A stronger model can often make sense of a vague request, but you’ll get better results by being specific — and by telling the model to ask you to clarify when it isn’t sure, rather than guessing.