Evaluate Agility with an AI Agent in 30 Minutes

If you already work with an AI assistant, it is the quickest way to get a feel for Agility. In about half an hour you can have the assistant answer your questions from the documentation, then connect it to a trial instance and watch it create a content model and content for you. Everything it saves lands in Staging, so nothing goes live without you.

What you need

• An AI client that supports MCP, such as Claude (Desktop or Code), ChatGPT, Cursor, VS Code with GitHub Copilot, Gemini or Windsurf.
• An Agility account and a trial instance. You'll create these in step 2 if you don't have them.
• About 30 minutes.

Step 1: Let the assistant read the docs (5 minutes)

The Agility Knowledgebase MCP Server gives your assistant search access to this documentation. It is public and read-only, and needs no sign-in.

In Claude Code:

claude mcp add agility-knowledgebase --transport http https://docs.agilitycms.com/docs/api/mcp

In Cursor, add it to .cursor/mcp.json:

{
  "mcpServers": {
    "agility-knowledgebase": {
      "url": "https://docs.agilitycms.com/docs/api/mcp"
    }
  }
}

Setup for Claude Desktop, ChatGPT, Gemini, Windsurf and other clients is in the Knowledgebase MCP Server article.

No MCP support? Give your assistant llms.txt, the index of every article. Any article URL with .md added returns clean Markdown, for example https://agilitycms.com/docs/overview/concepts.md.

Now ask the questions you'd ask a sales engineer. For example:

• "Using the Agility docs, explain the difference between content and pages in Agility."
• "How does preview work, and what do I have to build for it?"
• "How would I get all my content out of Agility if I left?"

Check the links it gives you. A good answer cites the articles it used.

Step 2: Create an account and a trial instance (5 minutes)

Go to app.agilitycms.com and sign up with GitHub, Google, Microsoft or an email address. Once you're signed in, create a new instance and choose a free trial. Quick Start: The Basics shows each screen.

Use a trial instance for this exercise, not one your team uses for real work.

Step 3: Connect the Agility CMS MCP Server (5 minutes)

The Agility CMS MCP Server lets your assistant read and change content in your instance. It signs you in with OAuth and acts with your own Agility permissions, so it can only do what your account can do.

In Claude Code:

claude mcp add --transport http "Agility-CMS" https://mcp.agilitycms.com/api/mcp

For other clients, use the one-click install buttons at mcp.agilitycms.com/instructions, or the per-client steps in the MCP server article.

The first time the assistant calls the server, a browser window opens so you can sign in to Agility. Then check that it's connected:

• "Which Agility instances can I access?"
• "Which user am I signed in as?"

Step 4: Build something by prompt (10 minutes)

Ask the assistant to model and fill a small slice of a real project. For example:

"In my trial Agility instance, create a content model called Event with a title, a date, a location, a short summary and an image. Create a content list for it, then add three sample events about our upcoming conferences."

Then follow up:

• "Show me the fields of the Event model you created."
• "Add a Speaker model and link each event to one or more speakers."
• "List the content items in the Events list."

While it works, watch what it does. It discovers the instance first, creates models and containers, and saves content items in batches. Each saved item comes back with a link you can open in Agility.

Step 5: Review it as an editor (5 minutes)

Open the instance in the Agility Content Manager and look at what the assistant made:

• The models and lists it created, under Models and Content.
• The content items, which are in Staging. Saves through the MCP server never go live by themselves.
• Publishing. Publish one item yourself. Publishing and deleting through the MCP server are gated: your client asks you to confirm before they run, and how it asks depends on the client. See Confirmation for destructive actions.

What to look for

Use this half hour to judge a few things about Agility, not just the assistant:

• Did the docs answer your questions, with links you could check?
• Did the content model make sense to someone who will edit in it?
• Did the safety model behave as described? Saves in Staging, your permissions applied, and confirmation before anything is published or deleted.
• How much would this shorten your project? If the answer is "a lot", the next step is a proper build.

Next steps

• Build a working site the same way. Vibe Coding with Agility and Building an Agility Site with AI Coding Tools take this from content to a deployed, editable site.
• Plan a real evaluation. Run a Two-Week Proof of Concept.
• Bring in your security team. Governing AI Access to Agility CMS explains authentication, permissions, confirmation and what leaves Agility when you use AI.