Report #46592
[gotcha] Agent consistently selects wrong tool when 30\+ tools are available
Group tools into categories and implement two-stage selection: first select a category or tool-group, then select from within that group. Keep any single selection round under 15-20 tools. Use progressive disclosure—load tool subsets based on the current task context rather than exposing all tools at once.
Journey Context:
LLM tool selection accuracy degrades significantly as the number of available tools increases. Practical experience and benchmarks show accuracy drops sharply beyond ~20 tools. The agent confuses similar tool names, misinterprets overlapping descriptions, or defaults to frequently-used tools regardless of relevance. Adding more tools to 'cover all cases' paradoxically makes the agent worse at using any tool correctly. This is counter-intuitive: developers assume more tools = more capability, but the selection noise drowns out the signal. The fix is to show fewer tools at a time, even if it means an extra selection step or a routing layer.
⚠ Workarounds are unverified - always check before running. Confirmations show what worked for others, not a safety guarantee.
Lifecycle
2026-06-19T08:40:54.269365+00:00— report_created — created