Search the web

Search normally

webctx search "golang http client retries"

webctx asks Brave, Tavily, and Exa, removes duplicate URLs, and returns one markdown list with short summaries.

When several providers independently surface the same page, that agreement helps the page rise in the final list.

Remove noisy sites

webctx search "react useEffect cleanup" --exclude medium.com,dev.to

webctx already excludes common video and social sites such as YouTube, TikTok, Instagram, and Facebook. --exclude adds your own domains for one search.

Look for a specific phrase

webctx search "drizzle orm" --keyword "migration guide"

--keyword uses Exa’s include-text search when you care more about a phrase appearing on the page than broad provider agreement.

Use the results with read-link

A good research loop is:

webctx search "OpenAI Apps SDK MCP annotations"
webctx read-link https://developers.openai.com/apps-sdk/reference

Search is for discovery. read-link is for turning the promising result into context an agent can actually use.

If you want the exact scoring model behind the merged list, see How search ranking works.

Output

Total Results: 12

- [Result title](https://example.com/page)
    - Short summary of the page

The final list is intentionally bounded so search does not flood an agent’s context.