Audit your store's SEO with Paperclip
Running a store SEO audit from Paperclip means assigning an agent on your org chart a task to read StoreSEO's product, collection, page and post scores, so the ranked worklist lands on the ticket as a written report.
Last updated: October 2026
Set up Paperclip for StoreSEO
Skip this if Paperclip is already connected. The full guide, with the alternatives, is on the Paperclip page.
- 1
Add StoreSEO to the agent Paperclip runs
Run this on the machine that hosts the agent, as the same OS user the Paperclip heartbeat runs as, from the agent's working directory. It is the claude_local adapter path, and user scope makes StoreSEO available to every Claude Code agent that user runs.
terminalclaude mcp add --transport http --scope user storeseo https://mcp.storeseo.com/mcp
- 2
Sign in once, then let Paperclip run headless
Start claude interactively as that same user, run /mcp, choose storeseo and finish the browser sign-in: pick your Shopify store and approve. Claude Code keeps the credentials, so later Paperclip runs need no browser. Check with claude mcp get storeseo.
new message to Paperclip/mcp
Step-by-step guide in the docs: How to connect StoreSEO MCP to Paperclip
How does this work in Paperclip?
You do not type into a StoreSEO window here. You create a task in Paperclip, assign it to an agent that already has StoreSEO on its runtime, and describe the audit in a sentence or two: which resource types, how many, and that it is a report. When the agent wakes, its runtime, for example Claude Code launched by the claude_local adapter, calls storeseo_start, which returns the conversation id every later call needs and a snapshot of how many resources are optimized, need improvement or are not optimized. It then calls storeseo_list_resources sorted lowest score first and storeseo_get_resource_seo for each candidate. Paperclip does not make those calls itself; it records the run, so you read the trace on the ticket.
What the agent posts back is the part Paperclip is good for. Each weak resource comes with its failing criteria in plain words, a GEO score for products, and a split between fixes a StoreSEO tool can make and fixes only you can make in the StoreSEO app, such as the focus keyword. Because the findings sit on a ticket, a teammate can comment on them, reassign the follow-up to another agent, or turn the top five items into separate tasks with their own budgets. Nothing in an audit writes to the store, which is why it is the safest first job to give a headless agent.
Once the one-off run reads well, put it on a rhythm. A Paperclip routine fires on a cron schedule and creates the same task each week, so the audit becomes a baseline you compare against the previous report. Keep the routine's wording explicit that it is read only, because the agent runs unattended.
-
You: Ticket: audit the store and post the ten worst products with what is wrong on each. Read only.
- storeseo_start read-only
- storeseo_list_resources read-only
- storeseo_get_resource_seo read-only
-
Paperclip: Posted on the ticket: the ten lowest products score between 19 and 36. Eight lack a meta description, six have images with no alt text, and two have no focus keyword, which you set in the StoreSEO app. No changes were made.
-
You: Open a follow-up task for the eight missing meta descriptions, assigned to me.
-
Paperclip: Created the follow-up task and left it unassigned to any agent, so nothing runs until you start it.
What to ask Paperclip
Tap a prompt to copy it, then paste it into Paperclip.
What StoreSEO does: store seo audit
StoreSEO scores every product, collection, page and blog post against its SEO checklist, and scores products for AI readiness (GEO) as well. An agent opens a session, lists the weakest resources first and reads the per-criterion analysis for each, so you get a ranked worklist instead of a dashboard to click through.
- 1
Open a StoreSEO session
The agent calls the start tool first. It returns a conversation id every later call needs, plus a snapshot of your store: how many resources are optimized, need improvement or are not optimized, image status, rank tracking and which features are set up.
- 2
List the weakest resources
It lists products, collections, pages or articles sorted by score, lowest first, or filtered to the ones marked not optimized. Products also carry a GEO score and their AI-readiness issues.
- 3
Read the checklist for each one
For each candidate it reads the full SEO record: current meta fields, focus keyword, score, every criterion that passes or fails, image status, FAQs and AI snippet.
- 4
Sort fixes by who can make them
It separates what a StoreSEO tool can fix (meta, handle, FAQs, AI snippet, image alt text, schema) from what only you can change in the app or in Shopify admin (the focus keyword, the product title, category, inventory).
- 5
Report a worklist
You get the resources in order, with the failing criteria in plain words and one suggested next fix. Nothing is written during an audit.
Store SEO audit reference
| Resource types | Products, collections, pages and blog articles |
|---|---|
| Scores | An SEO score for every resource, plus a GEO (AI readiness) score for products |
| Status labels | Optimized, needs improvement, not optimized; GEO adds not analysed |
| Sorting | By score, lowest first, or filtered to not optimized |
| Session | Every call carries the conversation id from the start tool; calls without it fail |
| What it writes | Nothing. Every audit tool is read-only |
| Fixed only in the app | Focus keyword choice, product title wording, Shopify category, product type, inventory, variant data |
| Plan | Any active StoreSEO plan, including the free plan |
Rules Paperclip follows
- Every audit call is a read and runs straight away; nothing in this job changes your store.
- The agent describes criteria by their visible title, never by internal identifiers or raw status values.
- Criteria no tool can fix, such as choosing the focus keyword, are handed to you with a link into the StoreSEO app rather than guessed.
- The agent should offer one next fix at a time, tied to what it just found, instead of dumping the whole list as a checklist.
In Paperclip
The audit tools are reads, but the agent that runs them is not limited to reads. The Claude Code adapter runs headless with dangerouslySkipPermissions on by default, so any StoreSEO tool that the model decides to call will run without a prompt. For a scheduled audit, say in the task text that it is report only, and consider setting dangerouslySkipPermissions to false or using the Paperclip connector with write actions on Ask first, so a stray preview or apply cannot run unattended.
What runs without asking. Reads such as SEO scores, store listings, settings and Google reports run straight away. Generating a preview spends AI credit but writes nothing, and StoreSEO tells the agent to say what a generation will cost before it runs one. Applying content, optimizing images, changing settings, publishing llms.txt or agents.md, writing a product description and deleting anything wait for your yes in the conversation. That yes is an instruction the agent follows, not a lock on the server, so your client's own tool-approval setting is the hard stop.
Questions
Can a Paperclip routine run the SEO audit every week?
Yes. A routine fires on a cron schedule or a webhook, creates a task and wakes the agent you assigned. Word the task as a read-only audit so the weekly run reads scores and posts findings without trying to change anything.
Where do I see what the agent asked StoreSEO?
Paperclip records the run against the task, so you can read the agent's tool calls and answers on the ticket. The calls themselves are made by the agent runtime, such as Claude Code, with StoreSEO attached to it.
Does an SEO audit change anything in my store?
No. The audit uses read-only tools: the session start, the resource list and the per-resource SEO record. Your store is only changed later, by a separate step you agree to.
What is the difference between the SEO score and the GEO score?
The SEO score measures a resource against StoreSEO's on-page checklist for search engines. The GEO score applies to products only and measures how ready the product is for AI shopping agents and generative search, using criteria such as FAQs, an AI snippet, alt text and product schema.
Can the agent choose my focus keyword?
No tool can set the focus keyword. The agent can tell you when the keyword is missing or looks wrong and give you a link to set it in the StoreSEO app, and several fixes such as keyword clusters and description writing need it set first.
How many products can the agent audit at once?
The listing is paginated, so the agent can walk your whole catalog page by page. In practice it is more useful to ask for the lowest scores first and work down from there.
Keep going
Store SEO audit with other agents
More Paperclip guides