Experience Modernization Agent FAQ faq

Short answers to questions that users of the Experience Modernization Agent often ask, from product fit and access through migration and continued development. Each answer links to the guide that covers it. For error messages and things that don’t work as expected, see Troubleshooting.

Product Fit product-fit

Do I need to be a developer to use the agent? audience

Not for a first result: Auto mode migrates your pages without step-by-step prompting. To change the site, you describe what you want in plain language, but you need some HTML and the basics of Git and GitHub, because changes reach production through pull requests. Authors edit content without Git. See Who Can Use the Agent.

Can several people work on the same project? collaboration

Yes, each in their own console workspace, sharing code through pull requests. The console doesn’t coordinate them, and the last upload of a page replaces earlier ones, so agree who works on what. See Work as a Team.

Can Edge Delivery Services run complex sites, or only simple ones? complex-sites

Complex enterprise sites. Pages are standard HTML, CSS, and JavaScript, so a block can call any API or run a complete application, and spreadsheets and the query index give blocks your own data. See How Your Site Is Built and Updated.

Is the agent limited to Edge Delivery Services? beyond-eds

No. Its skills focus on Edge Delivery Services, but it is a general-purpose coding agent: it can also adapt an existing application or write a server-side service for your team to deploy. For a system that it has no skills for, start by giving it that system’s documentation. See Work Beyond Edge Delivery Services.

Can the agent build integrations with other systems? integrations

Yes: the blocks and code that connect your site to commerce platforms, marketing technology, search, or your own APIs. Page import doesn’t carry the source’s scripts over, so plan each integration as project work. See Connect Services and Applications.

Can we teach the agent our own procedures? custom-skills

Yes. Write your conventions in AGENTS.md. For a recurring task that needs precise steps, add a project skill in .claude/skills/<skill-name>/SKILL.md; the agent reads it when a task matches its description. Its folder’s name starts with a dot, which changes how you push it. See Advanced: Custom Skills.

Can I change how the agent answers? reply-style

Yes. Write how you want answers in AGENTS.md, for example shorter answers, another language, or no suggested next steps. See Change How the Agent Answers.

How does the agent relate to Experience Workspace? experience-workspace

Experience Workspace is where authors edit the document-based content that the agent uploads; Edit in Experience Workspace in the console opens a page there. The agent itself works in the console, on the site’s code and on importing content. Experience Workspace has its own assistant for the content open in the editor. See Content Views.

Why can’t the agent use a capability that I’ve heard about? capability-availability

Some capabilities are optional add-ons, which are off until they’re turned on for the project. When your request needs one, the agent offers to turn it on. You can also ask it which add-ons are on, or ask for one by name. See Optional Capabilities.

Access and Usage access-usage

How do I get access to the console? get-access

Through your Adobe account team. You can sign up for Auto mode yourself at aemcoder.adobe.io/auto. See Access and AI Credit Usage.

I belong to several Adobe organizations. Which workspace do I get? organizations

One per organization. Projects and chats stay in the organization where you created them; to reach another organization’s, sign out and choose its profile when you sign in again. See Open the Experience Modernization Console.

Can I use a new or personal Adobe ID? adobe-id

Only if it belongs to your company’s Adobe organization. Your Adobe account team can add you to it. See Open the Experience Modernization Console.

If I can sign in to the console, is my organization licensed to use the agent? entitlement

Not necessarily. Your Adobe account team confirms your organization’s entitlement and commercial terms. See Access and AI Credit Usage.

Does the agent consume AI credits? ai-credits

Depending on your organization’s agreement, yes: for pages uploaded to your content destination, not for prompts. Auto mode runs don’t consume them. See Access and AI Credit Usage for what counts, and ask your Adobe account team about your organization’s usage.

Where can I see my AI credit usage? credit-usage-reporting

Contact your Adobe account team. An organization-wide total doesn’t identify how much was used by the Experience Modernization Agent. See Access and AI Credit Usage.

Can I use the console on a phone or tablet? mobile

Yes. The console adapts its layout to smaller screens, so that you can follow a long task on a phone. See Menu Bar.

Source Sites source-sites

Does it matter which CMS or framework the source site uses? source-technology

No. The agent works from the rendered page, not from the source’s code or CMS. Features that run on the source’s servers, such as personalization rules, search, or form processing, aren’t in the rendered page, so decide how each works on the new site. See Limitations.

Can I relaunch a site that runs on Adobe Experience Manager today? relaunch-aem-site

Yes. The agent treats it like any other source, and leaves the source site and your other Experience Manager sites unchanged. Components, experience fragments, and inherited content arrive as their output on each page. See Relaunch a Site That Runs on Experience Manager.

What if the source site uses bot protection? bot-protection

Arrange access with the site’s administrator, and don’t work around the protection. See Source Sites with Bot Protection for the options.

Can the agent migrate pages that require sign-in? protected-pages

Yes, when the source accepts one of the authentication methods that the console supports. Store a read-only credential in the console’s settings, or in Auto mode’s authentication fields, never in the chat. See Protected Site Authentication for the methods.

Can the agent reach a site behind a VPN, a firewall, or single sign-on? private-sources

Not through a VPN or an interactive sign-in. The agent reaches a source only over the internet, from its hosted environment, and it can’t complete an interactive sign-in such as single sign-on (SSO) or multi-factor authentication. A firewall lets it through when the site’s administrator allowlists the agent’s addresses. Otherwise, ask the administrator for a staging or origin host that the agent can reach; see Protected Site Authentication.

Auto Mode auto-mode

What does Auto mode do? auto-mode-what

It migrates pages of a website, including their design, without stopping for your review: it imports them, compares the result with the source, refines it, and emails you when it’s done. The guided workflow in the console stops after each step instead. See Start with Auto Mode.

Does Auto mode move my existing repository? auto-mode-repository

No. Auto mode migrates a website, starting from its URL or a list of page URLs. To work on a site that already runs on Edge Delivery Services, connect it in the console.

Can I guide how Auto mode identifies or builds blocks? auto-mode-guidance

Not during a run: you set its scope and options before it starts. To guide block identification or styling as you go, use the guided workflow, or refine the result in the console after the run.

How many pages does Auto mode migrate? auto-mode-pages

Up to 10 per migration by default, so a full-site run gives you a preview of the site, not all of it. Select Ask for a higher limit to request more, or import further pages in the console later. See Auto Mode Limits.

How long does an Auto mode run take? auto-mode-duration

Up to 24 hours. Checking and refining its own work makes it slower than a guided import of the same pages. See Start with Auto Mode.

Is the result of an Auto mode run ready for production? auto-mode-production

No. It’s a first result to review; Start with Auto Mode lists what to check. Continue in the console to import more pages and refine them, and for a release, follow the production migration guide.

What does Make it yours do? make-it-yours

It moves a finished run to a site that you own: the code to a new branch of your repository, and the pages and their media to your site’s content in Experience Workspace. Nothing is published or merged into main. See Make It Yours.

Do I need to create a repository before I use Make it yours? make-it-yours-repository

Yes. The destination is an Edge Delivery Services site of your own, whose content is stored in Experience Workspace. See Make It Yours for its requirements, and Create a New Site from the Boilerplate to set one up.

Does the transfer copy redirects, metadata, permissions, or site configuration? transfer-scope

No. Only code, pages, and the media that the pages reference are transferred. Set up spreadsheets such as redirects, bulk metadata, and placeholders, as well as permissions and configuration, in the destination. See Make It Yours.

Destinations and Git destinations

Can the agent write content to Experience Manager authoring with the Universal Editor? aem-authoring

Yes, in the guided workflow. The agent converts the pages to JCR content and uploads them to your Adobe Experience Manager as a Cloud Service author instance as a content package. See Getting Started for Experience Manager Authoring Projects. Auto mode and Make it yours produce document-based content for Experience Workspace only.

Which Git hosts can the console push to? git-hosts

GitHub, and any Git host that Cloud Manager connects, such as GitHub Enterprise, GitLab, Bitbucket, or Azure DevOps. See Use Another Git Host.

Do I need AEM Code Sync, AEM Code Connector, or a token? git-apps

AEM Code Sync on every GitHub project, so that your site can read its code, and AEM Code Connector for the console to push. A personal access token replaces the connector’s sign-in for GitHub Enterprise, or when OAuth sign-in isn’t available. See Start from a Repository.

Can the agent push to main or publish without my approval? approval

Only if you turn on its permissions in Settings > Agent permissions: with Git access, it can push even to main. Without them, code and content leave the console workspace only when you push, upload, or publish them yourself. To require approval, protect main and limit who has the site’s publish role, because the console’s own checks are safeguards, not approval gates. See Control over Changes.

Planning and Scale planning-scale

How do I use Plan mode? plan-mode

Switch the prompt input to Plan mode, with the light bulb icon, and describe your task. The agent proposes an approach and a checklist without changing files, and Execute plan carries it out. Plan mode works at any time, without a site catalog, and has nothing to do with the Planning view. See Prompt Input.

Why don’t I see the Planning view? planning-view

Planning is always in the view picker of the workspace panel, under Project; you don’t need to turn it on. It shows the site catalog, the inventory of templates, pages, and blocks that you review before migrating. Before you catalog a site, it asks for the site’s address. See Open the Planning View.

Can I switch between several migration plans? several-plans

No. The Planning view shows one plan per console workspace, so you can’t switch between plans, for example one per country site. Plan and migrate the sites one after the other. See Open the Planning View.

Can I choose which pages to migrate? choose-pages

Yes. In the console, name the pages to import, or choose templates and pages in the site catalog. In Auto mode, provide a list of page URLs instead of a site URL.

Can the agent migrate hundreds or thousands of pages? scale

Yes. The agent migrates a sample page for each template, and then generates import scripts that bring in the template’s other pages. Fixes go into those scripts and the blocks, so they reach every page of the template. See How the Agent Migrates at Scale.

How should we plan a large migration? plan-large-migration

Catalog the site, sign off a sample page for each template, and only then import the rest. For a site too large to catalog at once, the agent proposes how to divide it; see Invoking. A console workspace holds one site at a time, so in a program with several sites, push and upload each site’s work before you switch site. For assisted delivery, see AOE Delivery.

Blocks blocks

How should I ask the agent to build a block? block-prompt

Start with a short prompt that names the block and what authors put in it. The agent works out the content and the content model with you before it writes code. To match a design, give the page URL and the section, attach a screenshot, or select the element in the console preview with Inspect mode. See Prompt for a Block.

Can the agent reuse our existing blocks or block library? reuse-blocks

Yes. The agent first looks for a block to reuse or adapt in your project. During migration, it also checks the block library that Library URL in Settings points to. To reuse blocks from another repository, add them to this project first. See Reuse Existing Blocks.

Should I ask for many blocks in one prompt? many-blocks

No. The decisions that you make on the first blocks carry over to the next, so build them one at a time. See Build Blocks One at a Time.

Can the agent build a dynamic block after migration? dynamic-block

Yes. Ask for it like any other block, and describe the data source, what authors control, and what each item shows. The agent models the author settings as a key-value table. See Build a Dynamic Block.

Can I define how a block behaves before the migration? block-behavior

Yes: in the migration prompt, in the project instruction files for conventions that last, or in the Planning view before code is generated. See Set Block Behavior Before a Migration.

Where do form submissions go after migration? form-submissions

Wherever you choose. Edge Delivery Services doesn’t store submissions, and the source’s processing doesn’t come with the page. For organizations that license Adobe Experience Manager Forms, the forms add-on migrates each form to the Forms block, which by default sends submissions to the Forms Submission Service. For another system, such as a marketing automation platform, the agent builds the form and that platform’s administrator sets up the receiving side. See Form Submissions.

Does the commerce pages add-on work with platforms other than Adobe Commerce? commerce-platforms

It’s built for Adobe Commerce. It recognizes product pages on any source site, but the pages that it creates need Adobe Commerce as a Cloud Service or Adobe Commerce Optimizer, which can supply the catalog when another platform handles checkout. Without either, the agent can build product and listing blocks that call your platform’s API. See Commerce Pages.

Recovery recovery

What happens to my work if I reload the page or leave? reload

It stays: the chat history and the files in your console workspace are kept. If the workspace has paused in the meantime, see Open the Experience Modernization Console.

How do I restore a page that was overwritten? restore-page

Restore an earlier version from the document’s version history in Experience Workspace. A page that Make it yours overwrote has a version named Before Transfer. Code is recovered through Git. See Make It Yours.

Data and Security data-security

Can I give the agent a password or API key in the chat? secrets-in-chat

No. The chat reaches the AI model, and the agent can copy what you write into files that your site serves to every visitor. Enter credentials in the console’s Settings instead, and revoke any secret that you’ve already shared. See Keep Secrets Out of the Chat.

How can the agent call an API that needs a key while it works? api-credentials

Add the key in Settings > Protected site authentication as an entry for the API’s hostname, and tell the agent that it’s there. The console attaches it to the agent’s requests, so the AI model never sees it. See Credentials for Other Services.

Can the agent use its own token instead of my account? agent-tokens

Yes. Under Settings > Agent permissions, enter an Adobe service account’s token in Custom IMS Token and, where the Git permission is available, a scoped personal access token in Custom Git Token. See Settings Page.

Which AI model does the agent use? model

The agent currently uses Claude models. The model is an implementation detail that can change. See What the AI Model Processes.

Does the agent learn from my data? training

No. Adobe doesn’t train AI models on your data; see What the AI Model Processes. Between conversations, the agent knows only what is written in your project’s files; see What the Agent Remembers.

What data reaches the AI model, and which systems does the agent connect to? data-flows

The model receives your prompts, the source pages, the repository’s code, design references, and what the agent generates. The agent connects to your Git repository, your content destination, the source site, Figma when you use it, and the services that you set up credentials for, without receiving those credentials. See Security and Data Handling.

Is there an audit log of what the agent did? audit-trail

Not as a separate log, but the agent’s work is recorded. Everything that it produces reaches your site through Git, which records each code change, or through Experience Workspace or the author instance, which keep each page’s versions. To keep a conversation, select Download chat before you clear it. See Trace What the Agent Did.

Where can I find retention and compliance information? retention-compliance

From your Adobe account team and the Adobe Trust Center. For what the console itself keeps, see Where Your Work Is Stored.

Can I migrate a healthcare website? hipaa

Yes, with site content and fictitious data, not real patient information. That rule applies to public, staging, and signed-in pages. Your chosen authoring and delivery services have their own HIPAA requirements; see Health Data and HIPAA.

Help and Updates help-and-updates

How do I get help? help

Select Request support in the console, or Get help on the Auto mode pages; see Request Support for what to include. To meet experts or the product team, see The Team Behind the Agent.

Where are the release notes? release-notes

See the Experience Modernization Agent release notes. In the console, What’s new shows the highlights of each release.