Skip to content

What the MCP Tool Canvas is for

The MCP Tool Canvas is the AI 201 capstone. You design one tool on paper: what it does, where its data comes from, who may use it, what can go wrong, and how a human verifies the result.

Required fields

  • Tool Name, Purpose, Target User, Example User Question
  • Input Schema, Output Schema
  • Authoritative Source, Data Classification, Authentication Requirement
  • Permission Level, Risk Level, Human Confirmation Requirement
  • Provenance Strategy, Expected Errors, Fallback Behavior
  • Success Criteria

Example

search_courses

tool_name: search_courses purpose: Search official NTPU course information. example_question: Which generative AI courses are offered in 115-1? input: keyword = AI, semester = 115-1 authoritative_source: NTPU official course database data_classification: Public permission: Public Read risk: Low provenance: source, semester, retrieved_at, last_updated errors: semester_not_found, invalid_keyword, upstream_unavailable

A reminder before you submit

Warning

Do not enter secrets, tokens, credentials, personal data or confidential information. Use fictional or public examples only.

What may happen next

In the future, strong canvases may be reviewed for the NTPU AI Use Case Library, so that good designs can be shared and reused. This is not available yet.

Go to the canvas

What next?