Resources

Resources are the "read" half of MCP — data or documents an AI client can pull into context without calling an action.

What you can do

• Publish static resources: fixed content you write once, like a policy document, a glossary, or a config reference.
• Publish dynamic resources, backed by a live REST endpoint that Arthur calls at read time and returns as-is — with support for input defaults and an iterator path when the upstream response is a list that should become multiple resource entries.
• Declare a URI, an optional MIME type, and an error message shown to the AI client if the upstream call fails.
• Enable or disable each resource individually to control what's visible to AI clients and on the public documentation page.
• Write the resource's content as HTML (the default MIME type for a static resource is text/html) with a live Code/Preview editor, or, for a dynamic resource, render its live data through a Handlebars HTML template the same way a Tool's HTML output works.

How to use it

1. Open a server and go to its Resources tab.
2. Create a resource and choose static or dynamic.
3. For a static resource, write the content directly. For a dynamic one, point it at the endpoint that should back it, and set an iterator path if one upstream response should become several resource entries.
4. Enable the resource when it's ready for AI clients to read.

Good to know

• Resources are read-only by design — anything that changes data belongs in a Tool instead.