Some tools and data sources are enabled by being listed in the workflow and in each agent's tools list. They are disabled by omitting them from those lists (you can leave or remove the function definition in functions; if nothing references it, it won't be used).
Example: disabling the paper search tool
To disable the paper search tool (for example, to avoid Serper/API usage or to restrict agents to web search only), remove paper_search_tool from every tools list that references it. For example, in config_cli_default.yml you would:
- Remove
paper_search_toolfromintent_classifier.tools(orchestration node). - Remove
paper_search_toolfromdeep_research_agent.tools. - Optionally comment out or remove the
paper_search_toolfunction block infunctionsso the config is clearer.
Before (paper search enabled):
functions:
intent_classifier:
_type: intent_classifier
tools:
- web_search_tool
- paper_search_tool
deep_research_agent:
_type: deep_research_agent
tools:
- paper_search_tool
- advanced_web_search_toolAfter (paper search disabled):
functions:
intent_classifier:
_type: intent_classifier
tools:
- web_search_tool
deep_research_agent:
_type: deep_research_agent
tools:
- advanced_web_search_toolThe same idea applies to any other tool or data source: include it in the relevant tools lists to enable it, remove it to disable it.
For guidance on implementing and registering new tools or data sources, refer to:
- Adding a Tool -- How to create and register a new tool with the NeMo Agent Toolkit.
- Adding a Data Source -- How to add a new data source or knowledge layer backend.