How Can BigCommerce Partner Agencies Use AI Agents on Client Stores?
A BigCommerce partner agency can use AI agents on client stores by getting its own user account from the merchant, keeping each Stencil theme or Catalyst storefront in a repository the agency controls, and letting the agents draft changes for review. The agency checks every draft, the merchant approves it, and only then does it go live.
Last updated November 17, 2026
BigCommerce partner agencies are fielding the same question from clients and from their own teams: how do AI agents fit into work on a client's store without anyone losing control of the storefront? On BigCommerce the answer depends on three platform details: how user permissions work, whether the client runs a Stencil theme or a Catalyst headless storefront, and how checkout and third-party scripts are managed. This guide covers each, and which Noibu agent does what.
Key takeaways
- Ask the merchant for a named BigCommerce user account with only the permissions the work needs; high-risk permissions can only come from the store owner.
- Keep each client's Stencil theme or Catalyst app in a repository your agency controls, so agent drafts land on a branch, not the live storefront.
- Upload Stencil drafts without applying them, and review Catalyst changes on a staging deployment before anything reaches shoppers.
- Lead with the CRO agent on product pages and the Bug Resolution agent on theme, checkout and script code; the merchant approves every change.
- Treat checkout and Script Manager as their own lanes, with their own scopes and review steps.
Part of: What Is an AI-Native Ecommerce Agency?
Why are BigCommerce agencies bringing AI agents into client work?
BigCommerce's own podcast has made the case plainly. In the Keeping Commerce Weird episode The Agency Survival Playbook in the Age of AI, BigCommerce frames AI as raising the bar for agencies rather than replacing them, with focus and specialization as the way agencies stay valuable. For a BigCommerce partner, that means delivering more measurable improvement per client without adding headcount for every store.
AI agents help most with the work that never makes it onto a statement of work: product pages that underperform quietly, theme errors only some shoppers hit, and checkout problems nobody reports. A BigCommerce agency that runs agents on client stores can bring those findings to the client first, with a drafted change ready for review. The agency stays the expert in the loop, and the merchant keeps the final say.
What access does a BigCommerce agency need on a client store?
BigCommerce's user permissions guidance says each user should have their own account, with permissions limited to their areas of responsibility. For agency work, that means the merchant adds named agency users rather than sharing the owner login. A user who works on the storefront needs Store Design, which in turn requires Manage Channels, and anyone managing third-party code needs Script Manager.
Some permissions are deliberately harder to get. BigCommerce classes creating, listing and deleting store-level API accounts, and installing or launching apps, as high-risk permissions. Only the store owner can grant them, and both the owner and the user receiving them must use two-factor authentication. An agency should ask for high-risk permissions only when a task needs them, and should agree with the merchant up front who holds them.
BigCommerce's Agency Partner Program adds one more useful piece: partners can create sandbox stores in the Partner Portal. A sandbox gives an agency a safe place to test how a theme or checkout change behaves before it touches a client store.
Where should agent changes live on a Stencil theme?
Stencil theme code is a set of files that Stencil CLI bundles and uploads, which makes it a natural fit for a Git repository the agency's organization controls. When an AI agent drafts a change, it lands on a branch, and the agency's developers review the diff the same way they review any pull request. If the client or another vendor owns the repository, Noibu's guide to connecting AI agents to GitHub when you don’t own the repo covers that setup.
BigCommerce's Stencil deployment docs describe the detail that makes review safe. The stencil push command bundles and uploads the theme, then asks whether to apply it; any answer other than yes leaves it unapplied in My Themes, where the agency can preview it. The push needs a Stencil CLI token with the Themes: modify scope, which the merchant or a permitted user creates.
Two limits from the same deployment docs matter at agency scale. A store's My Themes section holds a maximum of 20 themes, and when a store is at that limit, Stencil CLI prompts you to pick themes to delete, with no further confirmation once you press Enter. The -d flag deletes the oldest theme automatically. Agencies that upload agent drafts often should name and prune review themes deliberately rather than letting automation clear them.
From agent finding to live BigCommerce storefront
- 1
Find
The CRO and Bug Resolution agents read the client store's sessions and errors and rank what to work on.
- 2
Draft on a branch
Changes land on a branch of the Stencil theme or Catalyst repository your agency controls.
- 3
Preview
Stencil drafts upload to My Themes without being applied; Catalyst drafts go to a staging deployment.
- 4
Agency review
Your developers check the diff, the preview and any checkout or script impact.
- 5
Merchant approval
Nothing ships until the client says yes.
- 6
Ship and measure
The change goes live and the agent reports what moved on the client store.
How does the workflow change on a Catalyst headless storefront?
BigCommerce describes Catalyst as a standard Next.js application built with React storefront components and the GraphQL Storefront API, and says it can be deployed to Vercel or any other provider that supports Next.js. For an agency, a Catalyst client looks like any modern web app: the code lives in a repository, and changes go through branches, review and a staging deployment.
That makes Catalyst a cleaner fit for agent work than many custom builds. An agent's change arrives as a pull request against the Catalyst app, the agency reviews it on a staging deployment, and the merchant approves it before it merges. One Catalyst difference to plan for: BigCommerce's scripts documentation notes that Catalyst does not support Handlebars, and that scripts load deferred through Next.js, so anything that relied on Stencil's head and footer script objects needs checking on Catalyst.
What about checkout and third-party scripts?
Checkout deserves its own review lane. BigCommerce's checkout customization docs list Optimized One-Page Checkout as the default, restyled through style overrides, with Open Checkout and the Checkout JS SDK for deeper changes on the storefront-hosted checkout. The same page notes that the SDK can filter shipping methods but cannot create new shipping or payment methods. Agents can find where shoppers fail at checkout and prepare changes, but every checkout change should be tested with real orders on a sandbox or staging store. For finding the cause of checkout problems first, see how to find the cause of checkout drop-off on BigCommerce.
Third-party scripts are where many client store problems start. BigCommerce's Scripts API docs say a script can target all pages, storefront pages only, checkout or order confirmation, and that adding a script to checkout requires the Modify Checkout Content scope and at least one Subresource Integrity hash to meet PCI DSS 4.0 requirement 6.4.3. The docs also cap control panel scripts at 50 per channel. Agents that trace an error or a slow page back to a script give the agency a specific line item to raise with the merchant, which is the same work covered in how to speed up a BigCommerce store without a developer.
| Store layer | Where the code lives | Where the agency reviews | Agents that work here |
|---|---|---|---|
| Stencil theme | Theme repository, pushed with Stencil CLI | Upload without applying, preview in My Themes, then apply | Bug Resolution agent (pull request), CRO agent (product page drafts) |
| Catalyst storefront | Next.js app repository | Staging deployment of the Catalyst app | Bug Resolution agent (pull request), Performance agent |
| Checkout | Optimized One-Page Checkout styles, or a forked Open Checkout build | Test orders on a sandbox or staging store | Bug Resolution agent, A/B Testing agent |
| Third-party scripts | Script Manager or the Scripts API | Check which pages the script loads on before enabling it | Performance agent, Bug Resolution agent |
Which Noibu agents do what on a BigCommerce client store?
For a BigCommerce agency, the CRO agent is usually the best first agent to put to work, because product pages are where clients feel conversion. The CRO agent reads the client store's session data in Noibu, finds the product pages that underperform, and rewrites them to convert better. The rewrite comes back as a draft: the agency reviews and edits it, the merchant approves it, and only then does it go into the client's product content or Stencil template. After the change is live, the agent measures whether the page converts better.
The Bug Resolution agent covers the code side. It identifies the top issues shoppers are hitting on the client store, traces the root cause to the codebase, whether that is the Stencil theme, the Catalyst app or a script, and creates a pull request for the agency to review. The merchant approves before the pull request merges and the theme is applied or the app deploys. Both agents follow the same loop: find, build, your approval, ship, measure, and compound, so every approved change makes the next one sharper.
Noibu runs six supervised agents in total: Bug Resolution, Performance, CRO, ADA, A/B Testing and ROAS, covering code issues, Core Web Vitals from real shopper traffic, product pages, accessibility audits against WCAG, multivariate experiments and paid traffic quality. Agencies can work with them in the Noibu console, and through Noibu MCP in AI tools their teams already use, such as Claude, ChatGPT, Microsoft Copilot and OpenAI Codex. Nothing ships without approval.
How does this differ from the Shopify agency workflow?
The approval model is the same on every platform: agents find and draft, the agency reviews, the merchant approves. What changes is the plumbing. Shopify partners work through collaborator accounts and Shopify's GitHub integration for themes, covered in how a Shopify agency can use AI agents on client stores. BigCommerce partners work through named store users, Stencil CLI or a Catalyst repository, and separate checkout and script lanes. Noibu works on any ecommerce platform, so an agency with clients on both can run one review process across its whole book, as described in how an agency can run AI agents across many client stores.
Agreeing who signs off is the part agencies most often leave vague. Settle it per client before the first agent draft arrives, using the options in who approves AI changes on a client’s store.
Where should a BigCommerce agency start?
Start with one client store and one agent. Get a named user account with the permissions the work needs, move the Stencil theme or Catalyst app into a repository your agency controls, and agree the approval path with the merchant. Then let the CRO agent produce its first product page drafts and the Bug Resolution agent its first pull requests. To see what agents would surface on a client store before any setup, run Noibu's free website audit on it.
Noibu is AI for ecommerce operations, the always-on optimization system that keeps online stores running at their best. A team of supervised AI agents finds what's broken, slow, or leaking revenue, does the work, and hands results back for human approval. Trusted by retailers like Free People, Guess, Liquid I.V., and Ariat.
Frequently asked questions
Do AI agents need the BigCommerce store owner's login?
No. BigCommerce recommends that each user has their own account with permissions limited to their responsibilities. The merchant adds the agency as a user, chooses its permissions, and the agents work through the theme or Catalyst repository and the tools the agency connects.
Can an agency push a Stencil theme to a client store without making it live?
Yes. Stencil CLI asks whether to apply a theme after it uploads, and any answer other than yes leaves it unapplied in My Themes. The agency can preview it there and apply it once the merchant approves.
Does this work on Catalyst headless storefronts?
Yes. Catalyst is a standard Next.js application built on BigCommerce's GraphQL Storefront API, so agent changes go through the app's repository and a staging deployment like any other code change before the merchant approves them.
Which Noibu agent should a BigCommerce agency start with?
A practical starting pair is the CRO agent, which finds underperforming product pages and rewrites them, and the Bug Resolution agent, which traces issues to the client's theme or app code and creates a pull request for review. Performance, ADA, A/B Testing and ROAS agents cover speed, accessibility, experiments and paid traffic.
Can AI agents change a BigCommerce checkout?
Agents can prepare checkout changes, but checkout is its own lane. Optimized One-Page Checkout is restyled through style overrides, deeper changes run through Open Checkout or the Checkout JS SDK, and every change should be tested with real orders on a sandbox or staging store before the merchant approves it.
Does Noibu only work on BigCommerce and Shopify?
No. Noibu works on any ecommerce platform. The BigCommerce-specific parts of this workflow are the user permissions, Stencil CLI, Catalyst and the checkout and script options described here.
