Skip to content

Default MCP Catalog — Search

Web search and content retrieval. Remote services (Tavily, Exa, Firecrawl, Jina AI, SerpAPI) expect a vendor API key via ${VAR} placeholders in the headers; the community stdio entry (Brave Search) needs BRAVE_API_KEY in the env section. All five remote services have a free tier sufficient for tutorial-scale use.

Entries (6)

Click any card to expand the full spec inline — transport (Streamable HTTP / STDIO), authentication shape (OAuth 2.1 / API key / Bearer / none), required environment variables, vendor URL or stdio command, and the upstream docs link.

Tavily 🔑

search · global ga

AI-optimised web search and answer engine for grounding LLMs with up-to-date facts and sources. Requires TAVILY_API_KEY.

Vendor   Tavily · T1 vendor

Auth        API key

Click for transport · auth · required env · description · docs
Exa 🔑

search · global ga

Neural web search by Exa — semantic ranking, source-aware retrieval, and high-quality result snippets. Requires EXA_API_KEY (x-api-key header).

Vendor   Exa · T1 vendor

Auth        API key

Click for transport · auth · required env · description · docs
Firecrawl 🔑

search · global ga

Web scraping, crawling, and structured data extraction with anti-bot handling. Bearer-authenticated with FIRECRAWL_API_KEY.

Vendor   Firecrawl · T1 vendor

Auth        Bearer

Click for transport · auth · required env · description · docs
Jina AI 🔑

search · global ga

Jina AI Reader (URL-to-clean-markdown) and search APIs for LLM-grounded retrieval. Bearer-authenticated with JINA_API_KEY.

Vendor   Jina AI · T2 vendor

Auth        Bearer

Click for transport · auth · required env · description · docs
SerpAPI 🔑

search · global ga

Real-time Google, Bing, Baidu, DuckDuckGo, Naver and other SERP scraping with structured JSON. API key embedded in URL path (SERPAPI_API_KEY).

Vendor   SerpAPI · T2 vendor

Auth        API key

Click for transport · auth · required env · description · docs

Workflow combinations

Five remote search services and one community stdio entry. They're substitutable but each has a different latency / quality / quota profile, so the meaningful "combination" is choosing one as primary and one as fallback:

  • Primary + fallbackTavily (LLM-tuned answers) primary, Exa (semantic search) fallback. Tavily fails fast on quota; Exa picks up.
  • Crawl + indexFirecrawl (crawls + structured extract) feeding a downstream Vector Database run. Good when the agent needs the content of pages, not just the URLs.
  • Search-engine resultsSerpAPI when you specifically want Google / Bing SERP JSON without scraping.
  • Long-form context retrievalJina AI for embedding + reader API; the model gets cleaned-up article bodies, not raw HTML.
  • Privacy-first local-stdioBrave Search (stdio) when you don't want a remote OAuth dance and you have a Brave API key.

Auth & secrets

All six entries are API-key gated. Set the key as a ${VAR} placeholder in the form (header for remote, env for stdio):

Connection Auth Env / header
Tavily API key Authorization: Bearer ${TAVILY_API_KEY}
Exa API key x-api-key: ${EXA_API_KEY}
Firecrawl Bearer Authorization: Bearer ${FIRECRAWL_API_KEY}
Jina AI Bearer Authorization: Bearer ${JINA_API_KEY}
SerpAPI API key (query param style; configurable as ${SERPAPI_KEY}) per-vendor docs
Brave Search (stdio) API key env.BRAVE_API_KEY = ${BRAVE_API_KEY}

Every remote vendor on this page ships a free tier large enough for tutorial-scale agent use — the free quotas are intended for evaluation, not production, but a single agent session won't exhaust them.

Picking guide

If you need… Reach for
LLM-shaped answers with citations Tavily
Neural / semantic search Exa
Crawl + structured extract Firecrawl
Embedding-friendly reader API Jina AI
Raw SERP JSON SerpAPI
Privacy-first, local-process search Brave Search (stdio)