AI feed management · Model Context Protocol
Fix your feed from your AI agent.
Your product feed breaks quietly — and you find out in Merchant Center days later. Connect Simple Product Feeds to your AI agent and diagnose, fix, and publish it in plain language. Preview-first, with a full audit trail.
Connect in under two minutes with an API key · read-only or read-write scopes.
The problem
The feed decides your Shopping performance — and no one watches it
A rule excludes too many products. A column mapping drifts. Google disapproves 600 items overnight. Nothing alerts you — you notice when the sales dip shows up, then lose an afternoon in Merchant Center guessing which of hundreds of errors actually matter.
The mechanism
Your whole feed toolkit, in the agent you already use
Simple Product Feeds exposes its full capability to your agent through a remote MCP server — 15 tools that cover the entire loop: inspect, explain, fix, publish, verify. No dashboards, no context-switching.
Preview-first by design
An agent cannot change a rule without showing you the diff first. The apply call requires a signed token proving the exact change was previewed — so nothing edits your live feed behind your back.
Every change is audited
Each mutation lands in a per-shop audit trail: which tool, which API key, what arguments, when. Overrides restore, mappings revert, rules deactivate — every change is reversible.
You: “Why did 40 products drop out of my Google feed?”
→ spf_feed_health(feed_1)
634 included · 40 excluded · GMC: 0 errors, 40 warnings
→ spf_debug_row(feed_1, row: 44821)
excluded by Rule #3: “Exclude when [D]inventory_quantity = 0”
You: “Keep out-of-stock items in the feed but mark availability.”
→ spf_preview_rule(…) — shows the diff on 40 sample rows
+ availability: out_of_stock
- excluded
Agent shows you the diff and waits for approval.
→ spf_apply_rule(preview_token) · spf_run_feed(regenerate)
✓ published · 674 included · re-validated clean
How it works
Connect in under two minutes
Create an API key in the app (Settings → API keys), then point your agent at the server. Works with Claude Code, Codex, Cursor — any MCP client.
$ claude mcp add --transport http simple-product-feeds \
https://app.simpleproductfeeds.com/v1/mcp \
--header "Authorization: Bearer $SPF_API_KEY"
Prefer project scope? Commit a .mcp.json — the key
reads from the environment, so no secret lands in git.
Frequently Asked Questions
- Is it safe to let an agent touch my live feed?
- Yes — that's the whole design. Writes are preview-first: the agent must show you the diff, and the apply call carries a signed token proving the exact change was previewed, so nothing edits your live feed behind your back. Every change is reversible, and every mutation is recorded in a per-shop audit trail. You can also issue a read-only key.
- Which agents work?
- Claude Code, Codex, and Cursor are documented, and any MCP-compatible client can connect. Agencies can use one org key across many client shops by adding a shop id to the connection URL.
- Do I need to be technical?
- No. You ask in plain language — "why did these products get disapproved," "fix my titles" — and the agent runs the tools, shows you the change, and publishes once you approve.
- What can it actually change?
- Rules, product overrides, column mappings, custom labels, and margin tiers — and it can publish the feed and sync channels. Every write is typed, validated, and undoable.
Stop finding out in Merchant Center.
Give your agent the tools to watch, fix, and publish your feed — safely.