Browse AI Agents & MCP SECTION 4: Model Context Protocol (MCP)

The MCP Server Designer

4 fill-in slots · from The AI Prompt Handbook for AI Agents & MCP

Your details

0/4 filled

The MCP Server Designer

Design an MCP server for this system.

System: [DESCRIBE WHAT IT IS AND WHAT IT DOES]
Who will use it: [WHICH AGENTS OR CLIENTS, FOR WHAT]
Data it holds: [DESCRIBE]
Operations it supports: [DESCRIBE]
Sensitivity: [PUBLIC / INTERNAL / REGULATED / PII]

Produce a design covering:
1. TOOLS — operations the model can invoke. For each: name,
   purpose, why it is a tool rather than a resource.
2. RESOURCES — data the model can read. For each: URI scheme, what
   it exposes, whether it is static or dynamic.
3. PROMPTS — reusable templates worth exposing to users. For each:
   name, arguments, what workflow it encodes.
4. What deliberately is NOT exposed, and why
5. The boundary between what belongs in the server and what belongs
   in the calling agent's own logic
6. Authentication and authorisation approach
7. What the server needs to know about the calling identity
8. Rate limiting and abuse considerations
9. Error taxonomy: what can go wrong and how each surfaces

Design principles to apply:
- Tools are for actions and computed answers; resources are for
  data the model reads
- Expose the task surface, not the internal API surface
- A server should be usable by a client that knows nothing about
  your internals

Do not write implementation code yet. Do not state specific
protocol version details — I will verify those against the current
specification.

Highlighted [slots] are still empty. Downloads are plain .txt — paste into any AI assistant.