Tool Diversity — Unique Tools Used

Tool Diversity counts how many distinct tools are used across all sessions in a day. It measures whether you and Claude are leveraging the full toolkit or relying on a narrow subset.

Definition

Tool Diversity is the count of unique tool names that appear in successful tool executions across all sessions for the day.

Formula:

Tool Diversity = COUNT(DISTINCT tool_name) from successful tool executions

Available tools

Claude Code has access to several specialised tools, each designed for a specific type of operation:

Tool	Purpose	Best for
Read	Read file contents	Viewing code, configuration, documentation
Edit	Make targeted, precise edits	Modifying specific lines or sections — more precise than rewriting entire files
Write	Create or fully rewrite files	New files, or complete file replacements where Edit would be impractical
Bash	Execute shell commands	Running tests, builds, git operations, system commands
Grep	Search file contents by pattern	Finding code patterns, function references, error messages across the codebase
Glob	Find files by name pattern	Locating files by extension, path pattern, or naming convention
Agent	Spawn a sub-agent for delegation	Breaking complex tasks into parallel subtasks handled by independent agents

Each tool fills a distinct role. Using only Read and Bash is like building only one unit type in a strategy game — it can work, but you're leaving capability on the table.

Data source

Tool Diversity is deterministic — counted from actual PostToolUse events:

• Each successful tool execution records the tool_name in plugin_usage_events
• The metric counts DISTINCT tool_name values across all events for the day
• Only successful executions count — failed tool uses (captured by PostToolUseFailure) are excluded

Interpretation

Diversity	Classification	What it means
1–2	Basic	Using a minimal subset. Often just Read + Bash. May be missing opportunities for more efficient approaches.
3–4	Moderate	Using a reasonable mix. Typical for focused sessions that don't need every tool.
5–6	High	Leveraging most of the toolkit. Indicates varied, complex work touching many aspects of the codebase.
7	Full	Using every available tool. Indicates a rich, varied session with delegation, search, editing, and execution.

Why it matters

Different tasks need different tools, and choosing the right tool makes work faster and more reliable:

• Edit vs Write: Edit makes surgical changes to specific lines without touching the rest of the file. Write replaces the entire file. Using Edit for small changes is safer and faster.
• Grep vs manual reading: Grep finds patterns across thousands of files in seconds. Reading files one by one to find something is orders of magnitude slower.
• Glob vs Bash find: Glob is optimised for file discovery within Claude Code. It's faster and produces better-formatted results than running find through Bash.
• Agent for delegation: Agent spawns sub-agents that can work on subtasks independently. Without Agent, complex multi-step tasks must be handled sequentially.

Low diversity isn't always bad — a session that only needs to read and edit files shouldn't use Bash just to increase the count. But consistently low diversity across many sessions may indicate that prompts aren't eliciting Claude's full capabilities, or that certain tools could make the workflow more efficient.

The StarCraft analogy

Tool Diversity maps to unit composition variety. A player who builds only marines can win some games, but a diverse army (marines, medics, siege tanks, science vessels) handles more situations and counters more strategies. Similarly, using a diverse set of tools means you have the right instrument for each task — precise edits when you need them, broad searches when exploring, and delegation when parallelism helps.