Artificial intelligence in the admin
Olive CMS includes artificial intelligence built directly into the admin area. It helps create content, translate text, and connect the website to AI assistants – without external tools and without copying content between windows.
Why use the AI features
- Content generation – instead of a blank page, you get a ready-made draft that you just adjust and approve.
- Translation in one click – one article, several languages, no manual copying of text between windows.
- Connecting AI assistants – Olive can be connected to tools like Claude or Copilot, which then work with the content directly in the CMS.
Generating content with AI
AI creates a draft of an entire page based on a short request – including structure, text, and images.
Three steps:
- You describe what you need. A short description is enough, for example "an article about the benefits of electric cars" or "a landing page for a new service." If needed, the AI will ask for more details.
- AI proposes a structure. Based on the request and existing content, it picks suitable content blocks – an intro section, text blocks, benefits, a gallery, or a call to action.
- You get a draft to approve. The generated content is saved as a draft. You edit it in the normal editor and publish it once you're happy with it.
What else the AI can do:
- Find a suitable image from a stock library for each section and upload it straight into the media library.
- "Learn" a style from an existing article you show it as an example, and create new content consistently with it.
- Automatically assign the article to the correct category (taxonomy).
Translating with AI
Content can be translated directly in the admin area, while preserving HTML formatting.
- Translation in one click – every article has a translate button; you pick the target language and the system handles the rest.
- Bulk translation – several articles can be selected and translated in the background at once, with progress tracked in real time.
- Preserved formatting – AI doesn't just translate text, it also works with the HTML structure, so headings and links stay intact.
- Shortcodes and tags – special markup in the text is detected automatically; if a translation could break it, the system warns you.
- Adjustable tone – translation rules (e.g. formality) can be set separately for each language.
- Ignored terms – product names, company terms, or abbreviations can be marked so the AI doesn't translate them.
As with content generation, translated content is saved as a draft, so there's always room for review before publishing. Categories and tags are translated along with the article, so the structure stays consistent across languages.
MCP – connecting AI assistants
Model Context Protocol (MCP) is an open standard that lets AI assistants (such as Claude or GitHub Copilot) work with the website's content directly, without copying text between windows.
What an AI assistant can do through MCP:
- Read and search content by type, language, or keywords.
- Create and edit pages – including structure, text, and images.
- Translate and localize content without losing formatting.
- Upload images from the stock library, from a URL, or from an AI generator.
- Work with the site's structure – it understands content types, content blocks, taxonomies, and the relationships between them.
How content is saved depends on the MCP tool, its parameters, and the connected account’s permissions. Some operations can also publish content. The MCP connection itself does not enforce human review before publishing. If content must remain a draft, check the selected tool’s draft setting and verify the content status in the administration area after completion.
Recommendations
- Review AI-generated and AI-translated content before publishing. When using MCP, first check whether the selected operation saves a draft or publishes content.
- For important or sensitive text, manually fine-tune the AI's result.
- If specific names or terms shouldn't be translated, add them to the list of ignored terms.
- If you're interested in connecting via MCP or in the AI features in general, contact your development/support team.