K:BOS 2026: What Klaviyo going headless means for you

Hey community! I watched the K:BOS 2026 keynote live so you didn't have to, and honestly, there's a lot to unpack. I've put together everything you actually need to know below, with my honest takes on what matters, what to try first, and what to watch out for. Grab a coffee and let's get into it. 

Jump to

Klaviyo announced six major areas of change at K:BOS 2026:

1. Headless Klaviyo — over 400 tools and nearly 500 APIs callable by Claude, ChatGPT, or a custom agent.

2. Natural-language SQL — ask questions about Klaviyo data through Composer or MCP and see the underlying query.

3. Personalization — Marketing Analytics becomes a broader decisioning layer for timing, channel, audience, and offer.

4. Composer — moves from a prompt-and-respond tool to one that runs recurring tasks, uses saved skills, and proposes edits for review.

5. Customer Agent — adds automatic language detection and support for 100+ languages.

6. New previews — Customer Hub expansion, Clicks to Bricks, and Compliance Hub.

The core change is not a new interface. It is that Klaviyo's data and actions are now reachable from tools your team already uses.

What does "headless" mean for a Klaviyo customer?

Headless means you can use Klaviyo's data and capabilities without opening the Klaviyo interface. Through a supported connector, you can ask Claude or ChatGPT to:

• Summarize campaign or flow performance

• Identify an audience segment that is changing behavior

• Answer a question about customer or event data

• Draft a campaign brief from recent performance

• Recommend what needs attention

• Take a supported action after you approve it

You can also build a custom workflow on the APIs or MCP: a reporting dashboard in another tool, a weekly performance summary posted to Slack, or a connection to your own agent.

Access is scoped, not universal. What an agent can see or do depends on the connector, your account role, the scopes you grant, and which tools are enabled. Headless access does not mean every AI tool can do everything the Klaviyo interface can do.

What is the Klaviyo SQL preview?

The SQL preview lets you ask data questions in plain language through Composer or MCP and get back both an answer and the SQL query that produced it. Initial coverage includes profiles, events, segments, lists, campaigns, and catalogs; forms, flows, and custom objects are being added next.

Klaviyo also introduced a semantic layer alongside the preview. This matters because terms like "revenue," "best customer," or "high-value segment" depend on your specific definitions, attribution windows, and exclusions. The semantic layer is meant to give Klaviyo and any connected agent a consistent definition of those terms; more customization for account-specific metrics is coming.

When an agent returns a number, confirm which data source, metric definition, and attribution window it used. An event-level total and a dashboard metric can answer different questions and produce different numbers.

What is changing with Personalization?

Marketing Analytics is being renamed and expanded into Personalization. This is more than a navigation change, but it does not replace every tool you use today.

Klaviyo is bringing existing personalization features into one decisioning layer that helps determine:

• Who to reach

• What to show them

• When to reach them

• Which channel to use

Named capabilities include Audience Optimization, 1:1 Content, Product Recommendations, Discount Decisioning, Personalized Send Time, and Channel Affinity. You still set the guardrails and the optimization target; the models operate inside those constraints, and holdout testing is intended to confirm the personalization is producing real lift.

Klaviyo states the platform processes more than 4 billion signals per day across billions of profiles. Because published profile-count figures vary across materials, confirm the current number against Klaviyo's latest product messaging before citing it externally.

What is new in Composer?

Composer is moving from a tool you prompt once to a tool that keeps working in the background. Depending on account rollout status, Composer can:

• Run scheduled or recurring tasks

• Save skills and connectors for reuse

• Analyze performance and explain what changed

• Create or iterate on segments

• Return citation-backed answers

• Turn a supported recommendation into a one-click edit

Composer can prepare or apply supported edits, but every edit requires your review and approval before it goes live. It does not independently send or publish campaigns. A Composer iOS app was also previewed, bringing questions and notifications outside the desktop interface.

What is new in Customer Agent?

Customer Agent now supports automatic language detection and response across supported channels, covering 100+ languages plus Shopify Markets localization.

Two capabilities were previewed: auto-training, which lets you refine agent behavior in plain language, and profile enrichment, which can write shopper preferences or intent back to the customer profile. This is distinct from using customer conversations to train a general-purpose large language model; Klaviyo's help documentation states customer data is not used to train large language models.

Voice was previewed as a future channel. Availability and pricing are not finalized, so treat it as an early preview rather than a generally available feature.

One published result: Klaviyo's Naked Wardrobe case study reports Customer Agent resolved 86% of customer queries and 94% of product-recommendation queries autonomously over a 90-day period. This is a single customer's result, not a universal benchmark.

What else was announced?

Customer Hub is live and expanding, positioned within Klaviyo's Marketing product line with additional on-site personalization functionality.

Clicks to Bricks is a preview for location-aware targeting, for example reaching shoppers near a physical store. Public detail is still limited.

Compliance Hub is coming soon, described as a centralized view of compliance health and requirements.

What can you use right now?

Some of these capabilities are still rolling out; help content may reference older status labels during that transition.

How do you connect an AI tool to Klaviyo safely?

Start read-only. The Klaviyo MCP server powers the Claude and ChatGPT connections; see also the Community's Claude and ChatGPT overview.

Four settings narrow what a connection can do:

• read-only=true disables write tools.

• disable-tools-with-user-generated-content=true reduces access to freeform customer text.

• core-tools-only=true reduces the overall tool surface.

• toolsets narrows exposed tools by API scope.

Read-only prevents changes but does not automatically hide profile data. For exact least privilege, such as read access to Lists and Segments with no Profile access, use a local MCP setup with a private API key scoped only to those permissions.

Klaviyo's current direction favors the remote server for most users; the local server suits more granular or backward-compatible setups. Review requested scopes before connecting. See How to restrict Claude's data access for common questions.

What should agencies do differently?

Use one clearly named remote connector per client. Klaviyo's MCP Server Guide for Agencies recommends this setup; connecting several accounts requires a paid Claude plan (Pro, Max, Team, or Enterprise) or ChatGPT developer mode.

This does not require a separate workflow per account. Klaviyo's agency guidance supports connecting multiple client accounts and querying across them from one place, provided each connector stays clearly named and scoped. For scheduled or cross-account automation needing explicit programmatic control, use the REST API with a separate private key per client.

Defaults worth following:

• Name every connector clearly by client.

• Start every connection in read-only mode.

• Review granted scopes before approving a connection.

Frequently asked questions

Is Klaviyo going headless?
Yes. As of K:BOS 2026, Klaviyo supports headless access: more than 400 tools and nearly 500 APIs can be called by Claude, ChatGPT, or a custom agent without opening the Klaviyo interface, subject to account role and connector scope.

What is the Klaviyo MCP server? The Klaviyo MCP server is the connector that lets Claude, ChatGPT, and custom agents access Klaviyo data and capabilities. It supports a remote server for most users and a local server for more granular, scoped access using a private API key.

Does Klaviyo use customer data to train AI models? No. Klaviyo's help documentation states that customer data is not used to train large language models. Features like Customer Agent auto-training and profile enrichment operate on an account's own data to refine that account's agent behavior, not to train a general-purpose model.

What is Klaviyo Composer? Composer is Klaviyo's AI assistant for analysis and campaign work. It can run scheduled tasks, save reusable skills, analyze performance, create or iterate on segments, and propose edits that a user must review and approve before they go live.

How much does Klaviyo Composer cost? Klaviyo's public pricing page lists a free tier with $5 of Composer usage per month. Some existing customers remain on earlier Early Adoption pricing. Check your account's pricing page for the number that applies to you, since published figures have changed across the transition.

What is the Klaviyo semantic layer? The semantic layer gives Klaviyo and any connected agent a consistent definition of business terms such as revenue, best customer, and high-value segment, based on an account's own attribution windows and exclusions, so natural-language SQL answers stay consistent with an account's existing metric definitions.

How do I restrict what an AI tool can access in Klaviyo? Start with a read-only connection. Use the read-only, disable-tools-with-user-generated-content, core-tools-only, and toolsets settings to narrow access, or use a local MCP server with a scoped private API key for exact least-privilege access such as Lists and Segments only.

Is Klaviyo's Personalization the same as Marketing Analytics? Personalization is the renamed and expanded version of Marketing Analytics. It keeps existing personalization capabilities and adds a broader decisioning layer covering audience, content, timing, and channel, while account owners still set the optimization guardrails.