AI assistants such as Claude or ChatGPT write texts, summarise sources and translate in seconds. Until recently, though, their help usually ended in the chat window: the result had to be copied, pasted into another system, formatted and saved. The Model Context Protocol (MCP) changes that. It lets an AI assistant work directly with your applications – for example with the content management system of your audioguide.
What is the Model Context Protocol?
MCP is an open standard that defines how AI assistants talk to external systems. An application provides a so-called MCP server, which describes the actions that are available – such as "create station", "update text" or "add image". The AI assistant can then use these actions in the conversation when you ask it to.
Think of MCP as a USB port for AI: one standard, many devices. Once connected, the same server works with different assistants, without a separate integration having to be built for every tool.
Why does this matter for museums?
In many museums, small teams look after a large amount of content: exhibition texts, audioguide stations, translations, the website, social media, press releases. Much of this work is transferring, adapting and repeating. This is exactly where MCP helps:
- Less copy and paste: the assistant writes content straight to the right place.
- Work in plain language: instead of filling in forms, you describe what should change.
- Many changes at once: shortening or harmonising twenty station texts is one instruction, not an afternoon.
- You stay in control: the assistant only works with the permissions of your own user account.
Managing the SmartCompanion audioguide via MCP
SmartCompanion offers its own MCP server that lets you manage your audioguide from an AI assistant. The assistant can read and edit your tours and stations, add images and audio files, adjust app texts and even change the branding of your app – logo, colours and images.
A few examples of what this looks like day to day:
- "Create a tour with eight stations for the new temporary exhibition. The texts are in the attached document."
- "Shorten all station texts of the 'Knights' Hall' tour to at most 120 words and keep the tone."
- "Add the English translations to all stations where they are still missing."
- "Replace the cover image of the family tour with the photo I just uploaded."
- "Change the primary colour of the app to match our new corporate design."
Larger files such as audio recordings or high-resolution photos are uploaded through a secure upload link that the assistant creates for you. This keeps the quality and metadata of your files intact.
Safe with drafts and publishing
Changes made via MCP first go into the draft of your audioguide. You can check them in the app preview before publishing them in the admin area. You sign in with your SmartCompanion account (OAuth) – there is no need to hand passwords or API keys to the assistant, and you can revoke access at any time.
MCP beyond the audioguide
The real strength of MCP shows when several systems come together. More and more applications offer MCP servers – from cloud storage and office tools to databases and content management systems. An assistant connected to several of them can take over workflows that used to be spread across many programs:
- From collection object to station: use object data and descriptions from a collection database as the basis for new audioguide texts.
- Accessible content: create texts in easy-to-read language or add image descriptions for visually impaired visitors.
- One source, many channels: turn one exhibition text into an audioguide station, a website article and a social media post.
- Check consistency: compare opening hours, names and dates across the website, audioguide and print material.
- Multilingual content: create translations and save them right where they are needed – read more in our article Translate your audioguide with AI.
What should museums keep in mind?
AI assistants are powerful, but not infallible. A few ground rules help you use MCP responsibly:
- Review the content: texts created by the assistant should be read by a knowledgeable person before publishing – especially historical facts.
- Grant access deliberately: only connect the systems a task needs, and use accounts with suitable permissions.
- Mind data protection: check which data your AI provider processes, and do not share personal data of visitors.
- Work with drafts: use preview and approval steps wherever your systems offer them.
How to get started
- Open the connector or MCP server settings of your AI assistant.
- Add the address of the SmartCompanion MCP server:
https://admin.smartcompanion.app/mcp - Sign in with your SmartCompanion account and confirm access.
- Ask the assistant, for example: "Show me all tours of my audioguide."
No SmartCompanion account yet? You can try SmartCompanion for free and test the MCP server with your own content right away.