Web search API for AI assistants

The web search API takes a query and returns pages from the open web that match, each with its title, URL, and a short snippet or description. Your assistant uses it to find pages, check facts, and pull links for you.

What it does

You pass a query, and you can filter by date range, language, country, or specific domains to include or exclude. Some providers let you pass a natural-language description of the page you want instead of keywords, and their index matches by meaning. The provider searches its index and returns the results, ranked by relevance or recency.

The reply has the URL, the page title, and a snippet. Some providers add metadata such as publish date, favicon, or structured highlights. A few providers support optional page content retrieval, which attaches the full text or a summary to each result for an added charge per page.

How a Joshu assistant uses it

When you ask Joshu a question it cannot answer from its training or your files, it searches the web first and reads the top results before replying.

When you are drafting a proposal or a pitch and want supporting data, Joshu searches for studies or recent articles, finds two or three good links, and puts them in the draft with a line on what each source says.

When you hand Joshu a list of company names for outreach, it looks up each one and flags the ones whose websites went down or changed industry, so you only send to the ones that make sense.

Choosing a provider

Providers differ in their indexing approach and result ranking. Keyword-based search engines return pages that match the exact terms you pass, while neural or semantic search engines let you phrase the query as a description of the kind of page you want and match by meaning. Some providers specialize in certain kinds of pages, such as research publications, companies, or people, and those specialized endpoints have their own capability pages.

Providers also differ in optional filters and in whether they support date or country parameters. A few offer different search depths or modes, with longer indexing or multi-query expansion at a higher price per call.

curl https://hello.joshu.me/api/router/call \
  -H "Authorization: Bearer $JOSHU_ROUTER_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "endpoint": "branddev.web.search",
    "params": {
      "query": "site:docs.context.dev Context.dev API",
      "numResults": 10
    }
  }'