Chapter C56 · Interactive Course

Tools, resources and prompts

Operations, addresses, templates, prepared workflows and errors.

Start chapter in web app

Core Learning Concept

Choose the protocol primitive that matches the need. Tools expose operations, resources provide addressable content, resource templates describe parameterized addresses, and prompts provide reusable messages. Follow listing, fetching, or calling as separate steps. Then inspect the result or error instead of assuming a valid request completed the task.

What You’ll Learn in This Chapter

  1. Tools offer callable operations

    Activity 1 of 8 · 4 min

    Identify an operation with arguments and a result.

    An MCP tool exposes a callable operation with an input schema and descriptive information. Examples include searching records or creating an item. Tool descriptions help selection but are not a complete audit of implementation. Inspect permissions and results, especially for operations that modify external state.

    Start activity →

  2. Resources expose addressable context

    Activity 2 of 8 · 4 min

    Recognize content available through a resource identifier.

    An MCP resource provides structured context at an identified URI, such as documentation or schema definitions. Reading resources supplies background knowledge but does not confer administrative authority or trigger server actions.

    Start activity →

  3. Resource templates describe a family

    Activity 3 of 8 · 4 min

    Explain a parameterized resource address.

    A resource template describes a family of resource addresses using parameters. For example, a URI pattern can represent notes identified by an ID. The template tells the client how addresses are formed; it does not necessarily enumerate valid IDs or grant access to every possible resource.

    Start activity →

  4. Prompts offer reusable message templates

    Activity 4 of 8 · 4 min

    Identify a server-provided starting workflow.

    An MCP prompt is a reusable message template offered by a server, potentially with arguments. A host can present it as a starting point for a task. Retrieving a prompt does not itself execute every described operation, and its contents do not override higher-priority host instructions or permissions.

    Start activity →

  5. Discovery lists what is offered

    Activity 5 of 8 · 4 min

    Distinguish listing capabilities from using them.

    Discovery lists capabilities or items that a server offers under the relevant protocol rules. Calling a tool or reading a resource is a later operation. A listed tool may still require authorization or fail on particular inputs. Keep “offered,” “selected,” “called,” and “completed” as distinct states.

    Start activity →

  6. A schema and a result answer different questions

    Activity 6 of 8 · 4 min

    Inspect input requirements separately from returned content.

    An input schema checks the shape and types of a request. The returned result describes the operation’s outcome, and the destination state may be needed to verify a side effect. A valid event-creation request does not prove the event exists until execution and its result have been inspected.

    Start activity →

  7. Errors are part of the interface

    Activity 7 of 8 · 4 min

    Distinguish protocol failure from an operation's reported failure.

    A protocol error concerns the exchange, while an operation can also return a domain-specific failure or an empty valid result. These outcomes require different repairs. Inspect the response fields and context: no matching records is not necessarily a broken connection, and malformed arguments should not be repaired by blind retries.

    Start activity →

  8. Choose the right primitive

    Activity 8 of 8 · 4 min

    Match search, an addressable handbook and a reusable review template to tools, resources and prompts, then inspect one failed operation.

    Start activity →