CJA MCP Read Only permission and Conversation Insights

Adobe made a new MCP Read Only permission item for the Customer Journey Analytics MCP server generally available on 6 October 2026, and followed it on 8 October 2026 with Conversation Insights, which puts chat data from LLM-powered experiences into Analysis Workspace. The first is an access-control change you can act on this week. The second is a new data source that needs planning before anyone switches it on.

Two other items are dated for later in October, and two earlier promises have slipped without a new date. This post sorts out which of these deserve your time.

Key takeaways

• MCP Read Only became generally available on 6 October 2026. The old MCP Access item is now named MCP Full Access, and people who hold it keep every tool.
• Conversation Insights reached general availability on 8 October 2026, after being planned for 22 September 2026. It covers prompts, responses and agent metadata collected through Web SDK.
• Do an access review first: anyone who only needs to look at data through the MCP server is a candidate for the read-only item.
• Automatic component descriptions are dated 28 October 2026. Adobe Brand Visibility integration is listed for October 2026 with no day given.
• Total population reporting and streaming schedule upload are postponed, with general availability shown as TBD. Leave them out of current plans.

What changed in the October 2026 Customer Journey Analytics release notes?

Four features are listed as new or updated, and two are listed as postponed. The notes were last updated on 7 October 2026, and Adobe says they change several times a month because releases are continuous. Treat the dates below as the state of the page on that day.

The notes also list fixes across areas such as Analysis Workspace, exports, data views, data ingestion and Report Builder, identified only by ticket numbers. If one of those ticket numbers matches a problem you reported, check it. Otherwise there is nothing to act on.

How does the MCP Read Only permission change access to the Customer Journey Analytics MCP server?

It splits one permission into two. Before, the MCP server permission was a single item called MCP Access. Now there are two: MCP Read Only, which gives access to all read-only tools without letting a user create projects, segments or calculated metrics, and MCP Full Access, which is the renamed original and keeps access to all tools, including those that create, change or delete components.

The practical point is the second half of that last sentence. Anyone who held MCP Access can now, by definition, change or delete components through the MCP server. In a small team that is fine. In a company where fifty people have Workspace access and a handful have built shared calculated metrics that executive dashboards depend on, it is a risk worth a closing look.

The notes say nothing about how tools are grouped, so do not assume details beyond what Adobe states. For the specifics of what each permission can do, read Adobe's Customer Journey Analytics MCP server documentation, which the release notes point to.

My position: this is the most useful item in the batch for most teams, because it costs almost nothing to adopt and reduces what a mistaken or over-eager AI client can change. If your organisation connects any AI assistant to the MCP server, start here.

What should an administrator do about MCP permissions this week?

Run a short review. It takes an afternoon, not a project.

1. List everyone who currently holds the permission now called MCP Full Access. Because the old item was renamed rather than removed, this is the same group as before.
2. Sort them by what they do with it. Analysts who ask questions of existing data through an AI client are read-only users. People who build and maintain components are not.
3. Move the read-only group to MCP Read Only through your normal permission process. Do this in a window where you can tell people, so nobody finds a creation tool suddenly missing in the middle of a task.
4. Keep Full Access for a named, short list and write down who owns each person's reason for having it.
5. Add the permission names to your onboarding checklist so new joiners get the narrower one by default.

What to ignore: the renaming itself. Nothing in the notes says existing users lose anything. You do not need to re-grant access to people who already have it. Also do not treat read-only as a substitute for data governance elsewhere. The notes describe a permission for the MCP server, nothing broader.

If you test the change, use one non-admin account. Assign MCP Read Only, try to create a segment or a calculated metric through the MCP server, and confirm it is refused while a read request works. Do that before you move the whole group.

What does Conversation Insights let you analyse in Analysis Workspace?

It brings unstructured chat data into Analysis Workspace so you can report on LLM-powered browsing and buying experiences across your properties. According to the notes, you can collect prompts, responses and agent metadata from conversational agents, whether your organisation's own agents or Adobe Brand Concierge, through Web SDK. You can then analyse intent, tone and sentiment, work at scale using your existing schema, datasets and data views, and connect agent interactions to wider customer journeys to measure effects on conversion and engagement.

The last part is the point. Many teams already log chat transcripts somewhere, in a support tool or a vendor dashboard, but those logs rarely sit next to journey data. A shopper who asks an agent about delivery times, then leaves, then returns through email and buys, is three disconnected records in most stacks. Adobe's stated aim here is to make that one journey.

A worked example. A retailer runs a shopping assistant on its product pages. Today the chat vendor reports volume and satisfaction scores. With the chat data in Customer Journey Analytics, the analytics lead can compare conversion for visitors who asked about returns against those who did not, and see which intents tend to precede a purchase and which precede an exit. Whether that comparison is meaningful depends on your data, but the question becomes askable.

The notes mark general availability as 8 October 2026 and give no rollout start. They do not mention limits, pricing or prerequisites beyond Web SDK collection and the existing schema, dataset and data view model. Check those on Adobe's Conversation Insights page before you commit a budget.

How should you prepare your data before turning on Conversation Insights?

Start from the data, not the feature. Chat content is different from clicks. People type names, order numbers and health or money details into a chat box without being asked to.

The notes do not describe how personal data inside prompts and responses is handled, so settle that with your privacy and legal contacts and in Adobe's documentation before collection starts. Then work through these:

• Inventory your agents. Which ones exist, who owns them, and which could send data through Web SDK? A custom agent built by one team may not be on the analytics team's radar at all.
• Decide what you need. Intent, tone and sentiment analysis are described in the notes. If you only need intent, ask whether you need to store every full response.
• Map to your schema. The notes say analysis uses your existing schema, datasets and data views, so decide where chat fields live and how they relate to the identity you already use for journeys.
• Plan a pilot property. One agent on one property, for a defined period, gives you something to check before it spreads.
• Write down the questions first. 'Do conversations that mention pricing convert differently?' is a question. 'Let's see what's in there' is not.

What not to do: switch it on across every property because it is now generally available. Unstructured text produces a lot of dimension values, and a messy first month is hard to clean up in a data view you also use for everything else.

What about automatic component descriptions and Adobe Brand Visibility?

Both are smaller, and both have caveats in the notes.

Automatic component descriptions are dated 28 October 2026. You will be able to generate descriptions for dimensions, metrics, calculated metrics, segments and date ranges, one at a time or many at once. Adobe frames this around helping Workspace users pick the right component in organisations with big libraries. If your library is full of items named 'Orders v2 FINAL' and 'Revenue (use this one)', this helps. But generated text describes what a component appears to be, and a person who knows what it is for should read it before it spreads. Assign a reviewer to each data view rather than bulk-approving. The notes say the documentation link is still to follow.

Adobe Brand Visibility integration is listed for October 2026 with no day given, so I will not give one either. The stated purpose is to connect Adobe Brand Visibility with your Customer Journey Analytics data so you can measure how AI-driven discovery turns into website engagement and business outcomes. The notes also say the documentation link is to follow. Until it exists, there is nothing to configure and nothing to evaluate beyond asking your Adobe account team what the prerequisites are.

Which features have been postponed, and should you plan around them?

No. Two items sit under postponed features, and neither has a new date.

Total population reporting would let you analyse and report on entities defined in profile and lookup datasets in a connection, going beyond time-based event series. It was originally planned for 22 September 2026, and general availability is now TBD. If your team has been waiting on it to count customers who have not generated events recently, keep waiting, and keep any workaround in place.

Streaming media schedule data upload, for tracking viewership of past live content such as FAST platforms, local streams and live sports, shows a rollout start of 29 October 2025 and general availability as TBD, originally planned for 29 October 2025. The notes describe the rollout start, but not what that means for access. If you work in streaming, ask Adobe directly rather than assuming the feature is open to you.

The lesson from Conversation Insights is relevant. It slipped from 22 September 2026 to 8 October 2026, and that was a short slip. Treat any date on this page as a plan, and hold project commitments until the feature is listed as available.

How do you roll this out and measure whether it worked?

Sequence the work by effort and risk.

Measurement is specific to each item. For permissions, count the number of people with Full Access before and after; fewer is the aim. For Conversation Insights, judge it by whether you can answer a business question you could not before, such as whether a particular intent relates to conversion, rather than by the number of reports built. For descriptions, ask whether new Workspace users pick the right component more often, which you can judge by fewer questions to the analytics team.

I have no benchmark for how much any of this will improve your numbers and will not invent one.

Where does an implementation partner help, and where can you do it yourselves?

You can do the permission review yourselves. It is administrative, small and reversible.

Conversation Insights is where outside help starts to pay off, particularly if your agents are custom-built. The hard parts are not in Analysis Workspace. They are in deciding what the agent sends, getting it through Web SDK in a form that fits your schema, joining it to existing identity, and agreeing with privacy teams what is collected. Those cross engineering, analytics and legal, and a partner who has done the schema and identity work on other implementations can save you a bad first design. If you are looking, the Customer Journey Analytics partners page lists firms to compare, and find partners lets you narrow by need.

If you do bring one in, brief them on the specific question, not on 'implement Conversation Insights'. The Customer Journey Analytics RFP template is a reasonable starting point for the scope, data and governance sections.

For the rest, wait. Component descriptions, Brand Visibility and the postponed items do not need a partner yet, because there is nothing stable to build.

Sources

• Adobe Experience League — Current Customer Journey Analytics release notes for October 2026, last updated 7 October 2026.