search

Available on the root (default user) and naive.forUser(id).

// Web search
await naive.search.web("naive agent governance", 10);

// Read one URL (optionally ask for a specific extraction)
await naive.search.readUrl("https://example.com/pricing", "the price table");

// Deep multi-source research
const quick = await naive.search.research("state of x402 adoption", "quick");
const deep  = await naive.search.research("state of x402 adoption", "thorough");

Method	HTTP	Notes
`web(query, count?)`	`POST …/search`	Real-time results.
`readUrl(url, extract?)`	`POST …/search/url`	Fetch + read one page. `extract` is a natural-language hint.
`research(query, depth?)`	`POST …/search/research`	`depth` is `"quick" \| "thorough" \| "exhaustive"`.

research changes shape with depth. "quick" answers inline; "thorough" and "exhaustive" are asynchronous and return a job_id — poll it through jobs. Code that reads a result field directly will find it missing on the deep paths.

Per-user and AccountKit-gated by the search primitive. Also exposed to agents via agentTools(). See the Search guide.

SDK

Sub-clients