One copilot. Every model and tool behind it.
Type or speak your mission. The Hub breaks it into steps, picks the best available model or tool for each, falls back when one is down, and stops for human approval before anything goes out.
Commerce Copilot
Routing path
"Route plan" shows which capability, model and tool each step uses. "Get real answer" sends your question and your imported data to the best connected AI.
Output
Performance Dashboard
Built only from data you've imported (files, screenshots, links, agents). Empty fields show "—". Nothing is estimated.
GMV over time
GMV by platform
Top products by GMV
By platform
Import / Export
Feed the Hub without waiting for platform APIs. Drop seller-center exports, screenshots, or links. You can also drop or paste (Ctrl+V) on any page of the Hub. You check every import before it's saved.
1 · Drop, browse or paste
CSV, XLSX, JSON exports · PNG/JPG screenshots. Click to browse.
Defaults for new imports
Export
Data is saved in this browser on this device. Export the Hub JSON to back it up or move it to another computer, then drop that file back in.
Saved data
Agent inbox · for Claude Cowork, Claude in Chrome, Grok or any browser agent
An agent can open your seller centers, export or screenshot the data, then drop it here. Or it can paste JSON into this box. Agents can also call window.MthryveHub.ingest(json) from the page.
Reports
Turn imported data into a client or executive report. The quick report is free and instant, using rules only. The AI report writes Performance, Insights, Problems, Opportunities, Actions and Next Plan from your numbers.
AI reports run through the router's best connected model. Claude via Netlify AI Gateway uses Netlify credits, a few cents per report.
Export
Gamma: copy the outline, then in Gamma choose Create → Paste in text. Each section becomes a card.
AI Model Router
No request goes blindly to one AI. Every task is split into steps. Each step is routed by capability, then scored on quality, speed, cost, context, reliability, availability and permission.
The routing chain
Selection score
Only tools that have the capability, are permitted, and are usable (Connected or Demo) can win. Claude isn't hard-wired to strategy or ChatGPT to content. Each wins when its profile scores best.
Fallback system
You'll see: "Primary production tool unavailable. Switching to available fallback."
Test one capability
| # | Tool | Status | Permitted | Score | Result |
|---|---|---|---|---|---|
| Pick a capability. | |||||
AI & Tools Control Center
Every connection, its real status, how it signs in, and whether the Hub may use it. AI model status is checked live against the server. Everything else you set here.
Hub settings
| Provider | Capability | Status | Authentication | Permitted | Fallback |
|---|
Add a model or tool
New providers plug into the registry without rebuilding the core.
Capability Registry
The router's source of truth: what each tool can do, its cost, speed, context and fallbacks. Nothing is hard-coded.
| Provider / Tool | Capabilities | Quality | Speed | Cost | Context | Inputs → Outputs | Reliability | Recommended tasks | Fallbacks |
|---|
MCP: dynamic capability discovery
When a provider has an MCP server, its tools are listed at connect time and written into this registry.
Pipelines
Creative (commerce data → production → results) and business intelligence (data → strategy → executive report).
Creative Commerce Pipeline
Business Intelligence Pipeline
Brief → Higgsfield prompt generator
Fill in the commerce requirements and the Hub writes the concept, shot list and video prompt for you.
Master Ecosystem
"I connect my shops, tell the system my mission, and the AI Commerce OS figures out what needs to happen, why, which agents and tools should do it, what humans need to approve, and how we measure whether it worked."