Servir un recurso de texto estático

The simplest resource is a static one: a fixed string your server returns every time. Great for help text, policies, or config notes.

You expose it by writing a Python function and tagging it with a resource URI. When a client reads that URI, your function runs and returns text.

The resource decorator registers the function under its URI. The string you pass becomes the address clients use to fetch the content.

For a text resource, just return a Python string. The SDK wraps it into the proper MCP read response so the client receives clean content.

Give the resource a clear description so the model knows when to read it. A docstring on the function is used as that description.

A static resource takes no parameters; it always yields the same body. If you need to vary output, that is a job for templates, covered later.

The return value can be assembled in Python before sending. The result is still static to the client, since the same call gives the same text.

Use a resource when the model only needs to read context. Tools are for actions; reading reference text is exactly what resources are for.

Many hosts let the user attach a resource to a chat, dropping its content into context. The model then reasons over that text alongside the prompt.

Static resources load fast since there is no work to do. Reserve them for small, stable text so you do not flood the model with tokens.

Run the server and use a client to list and read the URI. Seeing your exact string come back confirms the resource is wired correctly.

Choosing the right primitive for fixed text.

A static resource is a function under a URI that returns fixed text. Add a docstring description and keep the body small. 📄