max_uses | no | number | — | — | Maximum number of web_search invocations the agent may make within a single API request. Passed to Anthropic’s web_search_20260209 tool; blank uses Anthropic’s default. |
max_results | no | number | — | — | Target number of search results to return per web_search call (e.g. 5, 10, 20). Hint passed through to Anthropic; blank uses Anthropic defaults. |
topic | no | string | — | — | Topic hint (e.g. ‘news’, ‘finance’, ‘academic’) to guide search intent and result relevance. Passed to Anthropic; blank omits the hint. |
user_location | no | object | — | — | Geographic and timezone context object to improve result relevance for location-aware queries. All subfields optional; empty object means no location hint. Item fields below. |
allowed_domains | no | array of string | — | — | Array of domain names (e.g. [‘nytimes.com’, ‘bbc.com’]) to restrict search results to only these sources. Mutually exclusive with blocked_domains; blank allows all domains. |
blocked_domains | no | array of string | — | — | Array of domain names to exclude from search results. Mutually exclusive with allowed_domains; blank excludes nothing. Passed to Anthropic’s web_search_20260209 tool. |
max_content_tokens | no | number | — | — | Token limit on fetched page content when web_fetch is used to retrieve results from URLs. Applied per fetch; blank uses Anthropic defaults (typically ~8000 tokens). |