Just tell Claude or ChatGPT —
from analysis to data setup
Connect Laplace over MCP and use the AI chat you already have to ask about sales — and to hand off product matching, ontology sheet links, lookups, custom formulas, dashboards and alerts, all in plain words.
Listed on the official MCP Registryio.github.laplacetec/ecommerce-analyticsOne request, a finished report.
Ask Claude or ChatGPT the way you'd ask a teammate. The AI pulls what it needs from Laplace and writes it up as a report.
ClaudeIt doesn't stop at answers. Manage and set up by chat.
Hand the click-through-the-menus work to the AI. Using the tools Laplace provides, it looks things up, creates and edits for you.
Analysis
“Which 5 products had the highest contribution profit last month?”
- Sales, contribution profit and ad ROAS
- First vs repeat purchase, purchase cycle, cohorts
- Cross-sell and purchase paths
- Read values from saved dashboards and charts
Data management
“Match the unmatched products using the mapping sheet in Google Sheets”
- Store listing ↔ base product matching and rules
- Preview a Google Sheets mapping, then apply it
- Ad ↔ product links and ad matching rules
- Change history and restore points
Ontology
“Attach the influencer sheet and add a group-buy sales column per product”
- Build custom assets from Google Sheets or existing data
- Attach reference tables with LOOKUP
- Add calculated columns with custom formulas
- Aggregate, preview, publish and roll back
Dashboards & alerts
“Add this week's sales-by-channel chart to my report”
- Create and edit custom dashboard charts and tables
- Restore an earlier dashboard version
- Create, edit and toggle Slack alerts
- Leave memos on charts
Ontology setup, in a single request.
Attaching a sheet (lookup), adding calculated columns (custom formula), checking the result and publishing — the AI does it step by step. It shows the match rate and a preview first, so you only need to confirm.
The sheet has SKU, start date, end date and influencer columns. I'll join on SKU.
Lookup attached. 1,251 of 1,284 ordered products (97%) matched; 33 SKUs aren't in the sheet.
Added "Group-buy sales" and "Group-buy share = group-buy sales ÷ total sales" columns.
Here's a preview of the top products. Confirm and I'll publish it so you can use it in dashboards.
Try saying something like this.
| When you say | What the AI does |
|---|---|
| “Show me unmatched products and link them to base products with similar names” Look up unmatched → search candidates → save matches | Look up unmatched → search candidates → save matches |
| “If a Meta campaign name contains 'serum', link it to the serum product” Preview the rule (how many it hits) → save ad matching rule | Preview the rule (how many it hits) → save ad matching rule |
| “Attach the cost sheet and add a margin column by brand” Link sheet → lookup → custom formula → preview → publish | Link sheet → lookup → custom formula → preview → publish |
| “Put a contribution-profit chart with channel fees into my 'Weekly meeting' report” Check chart spec → create chart → add to report | Check chart spec → create chart → add to report |
| “Ping me on Slack if yesterday's sales drop more than 20% below normal” Preview alert condition → create alert | Preview alert condition → create alert |
| “Numbers look off after yesterday's matching change — what changed?” Look up change history and diff → find a restore point if needed | Look up change history and diff → find a restore point if needed |
Safe to hand off.
Your permissions, as is
The AI only sees and changes what the connected person's project role allows. If you can't see it, the AI won't answer with it.
Preview before, history after
Sheet matching and ontology changes show the result before they apply, and every change is logged so you can roll back.
Only what you ask
The AI makes only the changes you explicitly request. Big moves like deleting ontology objects stay in the web app.
Connect your AI in 3 steps.
Connect your stores and ad channels. You can try MCP during the free trial.
Add the Laplace MCP address (https://gateway.laplacetec.com/mcp/) in the connectors menu of Claude or ChatGPT.
Sign in and you're connected. Ask "Show my Laplace projects" to check, then go.
| AI tool | Connection | Notes |
|---|---|---|
| Claude (web · app) | Account sign-in | Easiest to set up — recommended to start |
| ChatGPT | Account sign-in | Turn on developer mode in settings, then add it |
| Perplexity | Account sign-in | Available on Pro and higher paid plans |
| Claude Code · Cursor · VS Code | API key | Developer tools — one command to register |
- Growth · Pro · Enterprise plans
- Free trial: up to 100 calls
- Not available on the Starter plan
Frequently asked questions, answered up front.
Still have questions? Contact us anytime — we will tailor the guidance to how you plan to adopt Laplace.
MCP (Model Context Protocol) is the standard way for AI assistants such as Claude and ChatGPT to use data and features from other services. Connect Laplace over MCP and you can query Laplace data from the chat — and hand off setup work like matching, ontology and dashboards in plain words.
An upload only shows the numbers at that moment, and nothing flows back into your system. The Laplace MCP queries live data that updates daily, on the same definitions Laplace uses (contribution profit, base products, customer identity), and the matches, columns and charts the AI creates are saved in Laplace.
The AI only makes changes you ask for, and sheet matching and ontology changes are previewed before they apply. Every change is kept in the change history so you can roll back.
Growth, Pro and Enterprise. During the free trial you get up to 100 calls; the Starter plan doesn't include MCP. Plans are checked against the project owner.
No customer personal data is queried over MCP. The AI works only with aggregated metrics and settings within the connected person's permissions, and you can disconnect anytime in Laplace settings.
Yes. It is listed on the official MCP Registry as io.github.laplacetec/ecommerce-analytics.