Your AI, with feed-safe permissions
Fix your product feed from the AI you already use.
Ask what is wrong, review a proposed rule against real products, and approve the next step from the same conversation. Read access cannot change data. Write access must be granted explicitly, and every action stays visible in SKULayer.
Your AI asks. SKULayer checks. You approve.
The Free plan is free forever and includes read-only agent access. Works with Claude and any MCP client.
What it looks like in practice
A real shape of conversation, from the first question to a running feed. Every number your AI quotes comes from your live audit and Google's own Merchant Center verdicts, not a guess.
You
What's wrong with the Springfields feed?
Your AI
Reading the latest audit. 23 findings: 1,140 products are missing a brand (warning), 11 titles exceed 150 characters (warning), 3 items have a sale price above the regular price (error). Google Merchant Center is also reporting 14 items disapproved for availability mismatch. Want me to draft a fix for the missing brands?
You
Yes, fill them from the product type where it starts with the manufacturer name.
Your AI
Drafted the rule and previewed it against your real items: 1,140 matched, 1,138 would change, no required fields emptied, no exclusions. Two items have no usable source value and stay as they are. Nothing is saved yet. Approve it?
You
Approve, and run the feed.
Your AI
Rule saved and the run is in progress. Your live feed stays exactly as it is until the new version is ready. I will tell you when it finishes.
Illustrative example using shipped tools. Numbers are fictional.
Built so you can say yes to this
Letting software near a live commercial feed is a trust decision. These are the guardrails, and they are the product, not a settings page.
Read-only by default
Access tokens are scoped. A read token can inspect feeds, audits, items, and Merchant Center issues, and nothing else. Write access is a separate, explicit grant.
Preview before every write
A rule proposed by an agent comes back as an impact report first: items matched, fields changed, exclusions, and warnings for risky changes. Your live feed only changes when a new version passes every check.
Everything on the record
Agent actions land in the same activity log as human ones, and agent-written rules are ordinary rules: readable, editable, and reversible in the app. Optional email notifications tell you whenever an agent changes something.
Revocable in one click
Every connection and token can be revoked from the account page. Revoke it and the agent is out, immediately, with nothing left behind.
Powered by a secure MCP connection with OAuth sign-in. Setup, supported clients, and the full tool list →
Agent access questions
Which AI assistants work with SKULayer?
Any MCP-capable client. Claude is the one we test most: connect it on the web or in the mobile app and your feed tools appear in the conversation, which is what makes the from-your-phone workflow real. Cursor, VS Code, and other MCP clients connect too; setup for each is in the docs.
Can the AI break my live feed?
No. Write access is a separate grant you approve, every proposed rule is previewed before it saves, and your live feed only changes when a new version passes every check. Any rule an agent saves can be removed or rolled back in the app.
Is this an autopilot?
No, and deliberately. The agent does the tedious parts: reading the audit, matching issues to products, drafting the rule, running the preview. The decision stays with you. Nothing changes your feed without an explicit approval somewhere in the loop.
What can a connected agent actually do?
Read your audit and Google's own Merchant Center verdicts, search your items, draft and preview a fix, and run the feed when you approve it. Every write is preview-first. The full tool list is in the docs.
Do I need to be a developer to set this up?
No. For Claude, it is add connector, paste the SKULayer URL, sign in, approve the scopes. Two minutes, no code. Other clients take a small config snippet, documented step by step in the docs.
Ask your feed what's wrong
Connect a feed on the Free plan, then connect your AI and ask. It reads the live audit straight from the same conversation, and on a paid plan it proposes fixes, previews them, and reads back Google's own verdicts.
Start free