5/ Security customization: remove tools the agent doesn't need, filter fields from responses, redact sensitive data, fix parameters to constant values. The fewer tools and the less data, the less the AI can get wrong.
6/ Semantic versioning: minor/patch updates applied automatically. Major versions require explicit opt-in. Test on staging before promoting to production.
Full docs: okt.to/lJEhFg
This is the entire setup for AI-native commercetools development:
/plugin install commercetools @ commercetools
skills + subagents + MCP, all in
then describe what you want to build and go
Full docs → okt.to/S07JeG
5/ The install is one command:
/plugin install commercetools @ commercetools
skills + subagents + MCP, all in. Then describe what you want to build and go: okt.to/YDM5Hd
2/ Generative UI: the interface composes itself around your request. Ask which products are about to sell out before the weekend and MosAIc builds the view that answers it - the at-risk items, days of cover remaining, a reorder control beside every one. Ask something different and the interface rearranges around that instead.
Put them together: intent in, interface out. the old model was navigate, click, wait, switch, repeat. The new model is one step.
MosAIc is in early access now. GA coming very soon!
okt.to/6IKVB7
commercetools for Builders turns a prompt into a live enterprise storefront in minutes. Watch it happen (full clip): okt.to/ACp23U
Then go build yours 😉
Here's what outcome-based commerce software actually looks like in practice 👇
"Set up a BOGO promo for skincare in DE and FR starting Friday"
(that's the input, one sentence)
What comes back: "the promotion is ready, two SKUs are low on stock in FR, here are your options."
No 5 tools and no manual coordination across systems. The user makes the decision that actually requires judgment. Everything else is handled.
That's MosAIc: okt.to/vM5XEl