tdqs / dimensions / tool-count-appropriateness
Tool Count Appropriateness
Is the surface well-scoped?
Summary
- level
- server coherence
- weight
- one quarter of the coherence score
- key
- tool_count_appropriateness
How it is graded
Is the number of tools appropriate for the server's purpose? An agent sees every tool at once, so a surface that is too wide costs context and selection accuracy, and one that is too thin leaves tasks it cannot complete.
Score anchors
- 5
- well-scoped, each tool earns its place (typically 3-15 tools)
- 4
- slightly over or under but reasonable
- 3
- borderline (1-2 tools feels thin, 16-25 feels heavy)
- 2
- too many (26+) or too few (1) for the apparent scope
- 1
- extreme mismatch (50+ tools, or a single trivial tool)
How to score well
- Each tool should earn its place. Fold variants that differ only by a parameter into one tool with an enum.
- A server with dozens of tools costs every agent context before it has done any work; a single trivial tool is not a server.
See also
The normative text is the specification; this page restates it. All dimensions: dimensions.