With Epicweb AI Assistant for Optimizely, editors can already work directly with AI inside Optimizely CMS.

They can use AI Chat to analyze content, generate text, improve SEO, work with translations, update properties and perform CMS-specific tasks without leaving the editorial interface.

But what if you want to use the same capabilities from outside Optimizely?

What if Microsoft Copilot, ChatGPT, Teams agents, Claude, Gemini, Zapier or your own enterprise AI agent should be able to work directly with your content in Optimizely?

That is what Epicweb AI Assistant Agent API makes possible.

Agent API extends the AI Assistant beyond the Optimizely editorial interface and makes its capabilities available to external AI agents and automated workflows.

The result is simple:

One AI Assistant for editors, AI agents and automated workflows.

Video: Agent API in action

AI inside Optimizely

The starting point is Epicweb AI Assistant.

Inside Optimizely, editors can interact with AI in the context of the CMS.

The AI Assistant can understand the current content, work with pages and blocks, analyze properties and use tools specifically developed for Optimizely.

This is important because a generic AI model does not automatically understand how your CMS works.

It does not know your content types, properties, editorial rules, permissions or custom business logic.

The AI Assistant provides that CMS-specific layer.

It gives AI the context and tools required to work with your actual Optimizely implementation.

For editors, this means AI becomes part of the normal CMS experience.

Agent API extends this outside the CMS

Agent API takes the same concept one step further.

Instead of requiring a user to open Optimizely and work with AI Chat, external AI agents can communicate with the AI Assistant programmatically.

The external agent can send an assignment such as:

Find all product pages where the SEO description is missing and create suitable descriptions.

Or:

Update the content for this product based on the latest information from our product system.

Or:

Compare the English and Swedish versions of this page and identify important differences.

The external agent does not need to know exactly how each operation should be performed in Optimizely.

It delegates the CMS-specific task to the AI Assistant.

The AI Assistant can then use its available context, instructions and tools to perform the work.

This creates an important separation.

The external AI agent understands the larger business task.

The AI Assistant understands how to perform that task inside Optimizely.

Use your organization's preferred AI platform

Organizations are increasingly building their own AI ecosystems.

Some have standardized on Microsoft Copilot.

Others use ChatGPT, Microsoft Teams agents, Claude, Gemini or their own internal AI platforms.

Agent API makes it possible to connect those platforms to Optimizely without rebuilding all CMS functionality for every new AI solution.

Examples include:

• Microsoft Copilot
• Microsoft Teams agents
• ChatGPT
• Claude
• Google Gemini
• Zapier and automation platforms
• Microsoft Foundry-based agents
• Optimizely agent platform
• Custom enterprise AI agents
• Internal business applications

The external platform becomes the orchestrator.

Epicweb AI Assistant becomes the Optimizely expert.

This makes Optimizely part of your wider AI architecture instead of creating another isolated AI environment.

Reuse the same AI tools

One of the biggest advantages of Agent API is that external agents can use the same AI Assistant tools that are available inside Optimizely.

A tool can give the AI Assistant the ability to perform specific operations.

For example:

• Find pages and blocks
• Read content
• Analyze properties
• Update content
• Generate SEO metadata
• Create content
• Work with translations
• Find language versions
• Compare content
• Analyze images
• Interact with Commerce content
• Call custom business functionality

Developers can also create their own tools for functionality that is specific to an organization.

This means you do not need to build the same integration several times.

A developer can create one tool for the AI Assistant.

That tool can then potentially be used by:

• An editor using AI Chat
• A Copilot agent
• A Teams agent
• ChatGPT
• An automated workflow
• A custom enterprise agent

This significantly reduces duplicated integration logic.

Keep the CMS intelligence inside the CMS

This is one of the most important architectural benefits.

Without Agent API, every external AI solution may need to learn how your Optimizely implementation works.

You might need to teach each agent:

• Which APIs to call
• How your content types are structured
• Which properties should be updated
• How language versions work
• Which fields are required
• Which business rules apply
• How content should be created
• How your organization wants AI-generated content to behave

That quickly becomes difficult to maintain.

With Agent API, much of that intelligence can stay inside the AI Assistant.

The external agent does not necessarily need to know that a certain page type has twelve properties, that one property contains structured content and that another property must follow a specific format.

The external agent can simply describe the desired result.

The AI Assistant handles the CMS-specific implementation.

This creates a cleaner architecture and makes it easier to introduce new AI platforms in the future.

Use the same instructions everywhere

Tools define what the AI can do.

Instructions define how the AI should do it.

Epicweb AI Assistant can contain instructions describing how AI should work with your content.

For example:

• Brand and tone of voice
• SEO requirements
• Accessibility requirements
• Translation rules
• Required properties
• Content type-specific rules
• Formatting requirements
• Product information guidelines
• Organization-specific terminology

These instructions can be used when editors work inside Optimizely.

With Agent API, the same rules can also guide external agents.

This means you do not have to duplicate your AI instructions across every platform.

Instead of configuring the same brand rules in Copilot, ChatGPT, Teams and several automation workflows, the relevant CMS instructions can remain close to the content.

That gives organizations a much more consistent way of working with AI.

Build automated content workflows

Agent API does not require a human to start every request.

It can also be used by automated processes.

This opens up an entirely new category of use cases.

For example, a scheduled process could scan your content and identify pages with missing SEO metadata.

The workflow could then ask the AI Assistant to:

1. Read the page.
2. Analyze the content.
3. Generate an SEO title and description.
4. Save the suggested changes.
5. Make them available for editorial review.

The same approach can be used for many other tasks.

SEO automation

Automatically identify and improve:

• Missing SEO titles
• Missing meta descriptions
• Poor metadata
• Inconsistent metadata
• Missing structured content

Content quality

Agents can continuously check for:

• Missing required properties
• Outdated content
• Inconsistent terminology
• Accessibility issues
• Missing alternative text
• Editorial guideline violations

Translation workflows

An agent could:

• Identify missing language versions
• Compare language versions
• Detect outdated translations
• Generate draft translations
• Prepare content for editor review

Product content

For Optimizely Commerce, an external workflow could detect product changes and ask the AI Assistant to update:

• Product descriptions
• Feature lists
• SEO metadata
• Campaign content
• Translations

Content enrichment

Existing pages could automatically be enriched with:

• Summaries
• FAQs
• SEO content
• Related information
• Structured content
• Alternative text

Connect Optimizely to larger enterprise workflows

Agent API becomes especially interesting when Optimizely is only one part of a larger business process.

Imagine a workflow such as:

PIM → AI Agent → Epicweb Agent API → Optimizely

A product is updated in a PIM system.

An enterprise agent detects the change.

The agent asks the AI Assistant to update the relevant product content in Optimizely.

The AI Assistant understands how that information should be represented in the CMS.

Or:

CRM → AI Agent → Agent API → Optimizely

A marketing workflow identifies a new customer segment.

An agent can ask the AI Assistant to prepare campaign content for selected pages.

Or:

Analytics → AI Agent → Agent API → Optimizely

An analytics system identifies poorly performing landing pages.

An agent asks the AI Assistant to analyze those pages and prepare improvement suggestions.

Agent API makes the CMS accessible as an intelligent part of a broader business workflow.

Use existing Optimizely permissions

Giving AI agents access to your CMS requires governance.

Agent API uses the existing Optimizely security model.

Each configured Agent API identity can be mapped to an Optimizely CMS user.

The agent then works with the permissions of that user.

This means your existing CMS configuration can control what an agent is allowed to access.

For example, you could create:

• An SEO agent that only has access to selected websites
• A translation agent with access to specific language branches
• A product agent with access to Commerce content
• A read-only analysis agent
• An agent for a specific market or business unit

This allows organizations to reuse familiar Optimizely concepts such as users, roles and content permissions instead of creating a completely separate authorization model for AI.

Keep editors in control

Automation does not have to mean automatic publishing.

A common Agent API workflow is to let AI prepare changes as drafts.

Editors can then review the changes inside Optimizely before publishing them.

This gives you the benefits of automation while keeping editorial governance.

For example, an agent could prepare hundreds of SEO improvements overnight.

The next morning, editors can review the suggested changes and decide what should be published.

This is often a much more useful model than trying to remove humans from the content process entirely.

AI handles repetitive work.

Editors remain responsible for the final content.

Maintain context across conversations

Agent API also supports conversational scenarios.

An external agent can continue working with the AI Assistant across several requests.

For example:

Find the product page for Product X.

Then:

Compare it with the Swedish version.

Then:

Update the English version based on the differences.

Then:

Prepare an SEO description as well.

By maintaining the same session, the AI Assistant can continue working within the same conversation context.

This is useful both for interactive external agents and more complex multi-step workflows.

Why not connect AI agents directly to Optimizely APIs?

You can.

But in many cases, Agent API provides a more maintainable approach.

A direct integration means the external AI agent may need to understand:

• Optimizely APIs
• Content structures
• Property types
• Content models
• Language handling
• Permissions
• Versioning
• Custom business rules
• Organization-specific functionality

You may also need to recreate the same logic every time another AI platform is introduced.

Agent API introduces an AI layer between the external agent and the CMS.

Instead of exposing all CMS complexity to every agent, you expose the AI Assistant.

The AI Assistant already knows how to work with the CMS and can use tools developed specifically for your solution.

This lets external agents operate at a higher level.

Instead of:

Call endpoint X, find content ID Y, update property Z with this value.

The request can be:

Update the SEO information for this product page based on the current content.

That is a significant difference.

A common AI layer for Optimizely

Organizations are likely to use more AI agents in the coming years, not fewer.

Different departments may use different platforms.

Marketing might use Copilot.

Developers might use ChatGPT.

Another team may build internal agents.

Automation platforms may run scheduled workflows.

Instead of building separate Optimizely integrations for all of them, Agent API provides a common entry point.

The architecture becomes:

Editors → AI Assistant

and

External AI Agents → Agent API → AI Assistant

Both can use the same:

• CMS context
• Tools
• Instructions
• Permissions
• AI configuration
• Business logic

This is the main idea behind Agent API.

One AI Assistant for both editors and AI agents.

Requirements

Agent API extends the existing Epicweb AI Assistant.

To use Agent API, your Optimizely solution needs:

• Epicweb AI Assistant base package
• AI Assistant Chat
• Agent API

The base AI Assistant and Chat provide the underlying conversational AI capabilities inside Optimizely.

Agent API then exposes those capabilities to external agents and automated workflows.

If you already use Epicweb AI Assistant and Chat, Agent API gives you a way to take the AI capabilities you have built inside Optimizely and connect them to the rest of your AI ecosystem.

Frequently Asked Questions

What is Epicweb AI Assistant Agent API?

Agent API allows external AI agents and automation platforms to communicate directly with Epicweb AI Assistant inside Optimizely CMS.

It extends the AI Assistant beyond the Optimizely editor interface.

What is the main benefit of Agent API?

The main benefit is reuse.

Editors and external AI agents can use the same AI Assistant, tools, instructions, permissions and CMS context.

You do not need to recreate your Optimizely AI integration separately for every AI platform.

What is the difference between AI Assistant Chat and Agent API?

AI Assistant Chat allows editors to work with AI directly inside Optimizely.

Agent API allows external systems and AI agents to communicate with the same AI Assistant from outside the CMS.

They serve different interfaces but use the same underlying AI capabilities.

Why should we use Agent API instead of connecting directly to Optimizely APIs?

A direct API integration usually requires the external agent to understand much more about your Optimizely implementation.

Agent API allows the AI Assistant to handle CMS-specific knowledge and tool execution.

This means the external agent can focus on the business task instead of the technical implementation of the CMS.

Which AI platforms can use Agent API?

Agent API can be used with platforms capable of communicating with an HTTP API.

Examples include:

• Microsoft Copilot
• Microsoft Teams agents
• ChatGPT
• Claude
• Google Gemini
• Zapier
• Custom enterprise agents
• Microsoft Foundry-based agents
• Other AI and automation platforms

Can Agent API be used for automation?

Yes.

Agent API can be called by scheduled jobs, background processes and automation workflows.

A human does not need to start every request.

Can external agents update content?

Yes.

If the required tools are available and the mapped CMS user has sufficient permissions, the AI Assistant can perform CMS operations on behalf of the external agent.

Does an AI agent get full access to Optimizely?

No.

Agent API can map an agent to an Optimizely CMS user.

The agent then follows that user's CMS permissions and content access.

Can different agents have different permissions?

Yes.

You can configure different identities for different agents.

For example, one agent could have access to marketing content while another only has access to a specific website or business area.

Can AI-generated changes be reviewed before publishing?

Yes.

A common workflow is to save AI-generated changes as drafts.

Editors can then review and publish them using the normal Optimizely editorial process.

Can we build our own AI tools?

Yes.

Developers can extend Epicweb AI Assistant with custom tools for your specific Optimizely implementation or business requirements.

Those tools can then be made available to editors and external agents.

Can we define how AI should work with our content?

Yes.

AI Assistant supports instructions that can define rules for tone of voice, SEO, translations, accessibility, content types and other organization-specific requirements.

The same instructions can then be reused when external agents access the CMS through Agent API.

Can Agent API be used with Optimizely Commerce?

Yes.

The same concept can be applied to Commerce scenarios, provided the required tools and functionality are available in your AI Assistant implementation.

External agents can then work with product and commerce-related content through the AI Assistant.

Do we need AI Assistant Chat to use Agent API?

Yes.

Agent API extends the existing Epicweb AI Assistant and Chat functionality.

You need the Epicweb AI Assistant base package together with AI Assistant Chat, plus the Agent API component.

Is Agent API only interesting for developers?

No.

Developers typically implement and configure the integration, but the value extends across the organization.

Editors can reduce repetitive work.

Content teams can automate quality and enrichment processes.

CIOs and digital leaders can connect Optimizely to broader enterprise AI initiatives without creating separate CMS integrations for every new AI platform.

Where can developers find the technical documentation?

The complete Agent API documentation is available in the Epicweb AI Assistant GitHub repository and includes configuration, authentication, sessions, OpenAPI support and implementation examples.