Brave's independent search API as an MCP server. Returns snippets and URLs the model can then fetch. The free tier is generous enough for personal research; paid tiers scale to millions of queries.

Web search results piped into the model, free tier 2k/month.

Use Case

Asking "what changed in TypeScript 5.7" and getting current results,Research on niche topics where Wikipedia or training data are stale,Citation hunting — get URLs to ground a claim,Competitive analysis: "compare pricing across X, Y, Z"

When Not to Use

When you need Google ranking specifically (Brave's index differs),For semantic search over your own docs (use Exa or a vector DB),High-volume scraping — they'll rate-limit you fast

Common Pitfalls

  • Free tier rate-limits at ~1 query/second; bursty workloads get 429
  • Image and news search are separate endpoints — default is web only
  • Some regional results differ from Google — verify if you need parity