Independent-index web search for assistants — fresh results without the Google API maze.
Brave Search runs one of the few genuinely independent web indexes left — tens of billions of pages crawled by Brave itself, not a re-skin of Bing — and its MCP server is one of the cleanest ways to give an assistant real web search. Ask something newer than the model's training cutoff and instead of a shrug or a hallucination you get actual results: titles, URLs, descriptions, freshness.
The official server covers more verticals than most people expect: standard web search, news with recency controls, image and video search, and local business lookups ("coffee shops near Waterloo station") backed by Brave's places data. Filters for country, language, SafeSearch and date range mean the assistant can construct reasonably precise queries rather than just keyword-blasting.
The pricing story is a big part of the appeal. The free tier — thousands of queries a month — comfortably covers personal assistant use, and the paid tiers are flat and predictable compared with the enterprise-sales gauntlet of some alternatives. You sign up at brave.com/search/api, grab a key, and you're searching within minutes.
Where does it sit against Exa and Perplexity's servers? Brave gives you classic search results — links to read, ranked by an independent index with no profiling. Exa returns semantically-matched sources tuned for AI consumption; Perplexity returns synthesized answers. For "find me pages about X and let my assistant read them", Brave is the straightforward, privacy-respecting default.
freshness explicitly (pd, pw, pm, or an explicit date range) when the question is time-sensitive — left unset, the ranker happily returns evergreen pages that predate whatever you're asking about.country and search_lang matter more than people expect for pricing, availability and regulatory questions: a US-defaulted query about an energy price cap will miss the UK answer entirely.brave_local_search falls back to ordinary web results when Brave has no places data for that area, so sanity-check the shape of what came back before trusting an "opening hours" answer.brave_summarizer is a Pro-plan feature and a second request — it consumes the summary key returned by a web search call rather than acting as a flag on the first one.Current events, new releases and recent docs pulled from a live index instead of guessed.
The assistant collects and cites URLs you can actually verify.
Places, hours and business info for 'near me' questions.
| Tool | What it does |
|---|---|
| brave_web_search | General web search with country, language and freshness filters |
| brave_news_search | Recent news results with recency controls |
| brave_image_search | Image search |
| brave_video_search | Video search |
| brave_local_search | Local businesses and places |
| brave_summarizer | AI-generated summary over search results (Pro plans) |
{
"mcpServers": {
"brave-search": {
"command": "npx",
"args": ["-y", "@brave/brave-search-mcp-server"],
"env": { "BRAVE_API_KEY": "your-brave-api-key" }
}
}
}Free API key from brave.com/search/api.
A Brave Search API key — the free tier's monthly quota covers typical personal assistant usage.
| Variable | Description | Required |
|---|---|---|
| BRAVE_API_KEY | API key from the Brave Search API dashboarde.g. BSA... | Yes |
The simplest web tool that matters — fetch any URL and get model-ready markdown back.
Industrial-strength web extraction — render, scrape, crawl and search entire sites into clean markdown.
Puppeteer-powered browser control that drives pages from the accessibility tree instead of pixels.
Search built for AI, not humans — semantic web search that returns model-ready content, plus code context.
Answers, not links — delegate questions to Perplexity's search-grounded models and get cited responses back.
Put 6,000+ pre-built scrapers at your assistant's fingertips through one MCP endpoint.