MCP · Web

Brave Search

Real-time web search via Brave API. Private alternative to WebSearch. Useful when your context needs fresh info (news, prices, competitors).

Official docs ↗Free up to 2000 queries/month

Use cases

›This week's news on a topic
›Up-to-date competitor pricing/features
›Fact-check a quote or stat

Installation

1.Create an account on api.search.brave.com. Free plan up to 2,000 queries/month, enough for heavy personal use.

2.One line, one terminal

$ claude mcp add brave -- npx -y brave-search-mcp --key=YOUR_API_KEY

Need to install Claude Code first? See the Claude Code page · Official docs

Concrete examples

›"Qu'est-ce qui s'est passé cette semaine dans l'UCaaS européen ? 3 infos majeures max."

Real-time web search + relevance sort. Useful to prep a brief or quick post.

Sample output

→ Brave search (window: 7d, topic: UCaaS Europe)

3 key updates

1. RingCentral announced closing its Paris office
   Source: Les Échos, Apr 12. 35 employees impacted, their FR market
   share stays. Signal: consolidation underway.

2. Wildix acquires Enreach Partner UK
   Source: Silicon.co.uk, Apr 14. Enreach exits UK, Wildix strengthens
   its position. Deal at ~£80m per unofficial sources.

3. Arcep publishes its report on enterprise VoIP quality
   Source: Arcep.fr, Apr 15. 3 operators flagged for peak-hour service
   degradation. The report can serve as a counter-argument to those who
   question UCaaS reliability vs physical PBX.

Items 2 and 3 deserve a post or an editorial angle.

›"Vérifie la stat : 'les posts LinkedIn avec question augmentent l'engagement de 50%'. Source et fiabilité ?"

Fact-checking. Claude searches the primary source, evaluates reliability, flags if the stat is repeated without source.

Sample output

→ Brave search: "LinkedIn posts question engagement 50%"

Fact-check result: UNRELIABLE

- The stat has circulated since 2019 in SEO articles ("10 tips for LinkedIn
  engagement...")
- No primary source found: not LinkedIn official, not academic research,
  not credible analytics tools (Hootsuite, Sprout, Buffer publish
  different stats)
- Closest study: Buffer 2023, 1.3M posts analyzed. Posts with questions
  have +22% engagement (not 50%).

Recommendation: don't use "50%" in your content. Use "~20%" with a Buffer
reference, or a less precise phrasing ("posts with questions perform
better").

Read the MCP article →