Failure Index / Agent behavior · warning · LangChain · OpenAI Agents SDK · CrewAI

Agent calls a tool that doesn't exist (hallucinated tool name)

The model invented a tool name — asked for search_web when the toolbox has web_search, or dreamed up a tool that sounds like it should exist. The framework can't dispatch it and the step fails.

The error

agent tool not found error
langchain agent invalid tool name
model called nonexistent function

Root cause

Tool names/descriptions are ambiguous or overlapping, there are too many tools in context, or the prompt references capabilities that were never registered. Models generalize: if the naming pattern suggests a tool should exist, they will try to call it.

The fix

Rename tools to be unambiguous and consistent (one naming convention), trim the toolbox to what the task needs, and make sure the system prompt never mentions capabilities that aren't actually registered as tools.

Preventing it next time

Keep toolboxes small per agent (5-10 well-named tools beat 40 vague ones) and validate tool-call names with a fuzzy-match fallback that corrects near-misses.

Stop debugging this by hand. Vorlo watches your agent and, when this failure happens, hands you the diagnosis and the fix — verified by developers who hit it before you — in your dashboard, Slack, or your editor. Two lines of code: pip install vorlo-trace · npm install vorlo-trace. Start free

Related failures

Agent stopped due to iteration limit or time limitAgent stopped due to iteration limit or time limitContext length exceeded: agent run grows past the model's windowcontext_length_exceeded agent