Ops: the AI agent that keeps the store fast and honest
Performance, caching, bot protection, and the shared record of everything the other four agents have done. Unglamorous work, done continuously.
Every conversion improvement is undone by a slow page, and every automated system is dangerous without an audit trail. Ops covers both. It runs on Cloudflare for edge performance, caching, and bot protection, and on Supabase as the shared memory layer where every agent reads context and logs what it changed. It is the least visible agent and the reason the others can be trusted.
Performance as a continuous job
Core Web Vitals degrade quietly — an app gets installed, an image goes unoptimised, a third-party script is added for a campaign and never removed. Ops watches the real-user metrics rather than a synthetic score, finds what regressed, and either fixes it at the edge or hands a page-level fix to Builder.
Real-user Core Web Vitals monitoring, segmented by device and template
Cache rules and edge configuration tuned per route
Third-party script auditing — what was added, what it costs, what still earns its place
Image and asset delivery optimisation across the catalogue
Security and traffic hygiene
Bot traffic distorts analytics before it does anything worse. Ops manages bot protection and rate limiting at the edge, keeps legitimate crawlers — including AI crawlers you want indexing you — unblocked, and reports on traffic that is inflating your numbers.
The audit layer that makes automation safe
Every action by every agent is written to a shared log: what was changed, by which agent, on what evidence, and with what result. That record is what lets you hand real work to software. Ops maintains it, keeps the schema consistent, and makes sure the other agents are reading the same version of the truth.
How it works
Connect Cloudflare and Supabase
Ops takes over edge configuration and provisions the shared memory and audit schema the whole team writes to.
Baseline the store
It measures current real-user performance per template and records the starting configuration so every later change is comparable.
Maintain continuously
Regressions are caught against the baseline, fixed at the edge where possible, and escalated to Builder when they are a page problem.
What you get
Core Web Vitals tracked against real users, not lab scores
A complete, queryable log of every automated change on the store
Bot and scraper traffic separated out before it corrupts your analytics
Edge caching tuned per route instead of one blanket rule
Handled by these agents
Frequently asked questions
Do I need to move my DNS to Cloudflare?
For edge performance and bot protection, yes — those run at the network layer. If you would rather not, Ops still maintains the shared audit layer and performance monitoring, with a smaller set of levers to pull.
What exactly is stored in the shared memory layer?
Store configuration state, agent findings, task history, and a log of every change with its inputs and rationale. It holds operational data, not customer personal data.
Will bot protection block AI crawlers?
Not the ones you want. Being retrievable by AI engines is a growth channel, so Ops keeps those crawlers allowed by default and blocks scrapers and abusive traffic. You control the list.
Can I export the audit log?
Yes. It is your data in your Supabase project, queryable directly, and exportable at any time.
Keep reading
Site speed optimisation
Find what actually slows real users down, fix it, and stop the regressions that undo it.
Improve store page speed
A field-data-first speed programme: measure real users, fix by layer, prevent regression.
Run a technical SEO audit
The store-specific audit: crawl reality, indexation truth, structured data, and speed.
Builder
Ships storefront, landing page, and product page changes without waiting on a dev sprint.
Put the team to work on your store
Connect your storefront, analytics, and email stack, set a goal, and let the agents run the work end to end. Start on the free plan with your own model key.