You configure routing rules in your agent's gambit logic. Common triggers include questions about current events, recent product releases, competitor information, or any topic where freshness matters. The agent calls You.com Search when the question falls outside its training data or when live verification is valuable.

You control this with the num_web_results parameter, which accepts values up to 20. For quick factual questions, 3 to 5 results usually suffice. For comprehensive research, set it higher. The agent synthesizes the results into a single conversational answer.

Tars requires your You.com API key, which you obtain from your You.com developer account. The API key authenticates all search requests. Usage limits depend on your You.com plan. You can monitor usage and rotate keys from the You.com dashboard.

No. Search results are fetched in real time during conversations and used only to generate the current response. Tars does not cache or store You.com search results. Each query produces fresh results from the live web.

Yes. The agent passes a country code parameter (like US, GB, DE, JP) to You.com's API. This localizes results to the specified region, ensuring customers receive geographically relevant information. Over 30 country codes are supported.

A regular search engine returns a list of links for the user to browse. Tars combines You.com search results with AI language capabilities to deliver synthesized, conversational answers. The customer gets a direct answer, not 10 blue links to sift through.

Yes. The agent supports safesearch levels of off, moderate, and strict. Set this in your tool configuration so every search respects your content policy. Moderate is the default, which filters explicit content while keeping results comprehensive.

The agent handles empty results gracefully by relying on its own training knowledge to answer the question, or by informing the customer that specific current data was not found and offering to help in another way. No dead-end responses.