Exa
The search engine built for AIs — semantic web search, page contents, Websets, and research APIs, plus the ecosystem's most-used search MCP server.
Exa is a search engine designed for AI consumers, not human browsers: a semantic Search API with deep-search profiles, a Contents API returning clean page text and summaries, Websets for building enriched entity sets, and research endpoints. Its hosted MCP server (mcp.exa.ai/mcp) is the most-used search server in the ecosystem and even works keyless on a rate-limited free tier.
Exa is what search looks like when the customer is an agent: semantic search in, clean text out. Where Google optimizes for a human scanning ten blue links, Exa's Search API returns machine-ranked results and its Contents API hands back the page as clean text, highlights, or AI summaries — the retrieval layer for agents and RAG pipelines, sold as an API.
Highlights
- Search built for LLM consumption — neural/semantic search with speed/quality profiles up to Deep Search and deep-reasoning modes for research-grade queries.
- Contents, not links — clean page text, highlights, and summaries per result; structured outputs via an
output_schemaparameter. - Websets — build and enrich entity sets ("every Series-B devtools company and their CTOs") as a product, not a scraping project.
- Research & Monitors — multi-step research runs and standing watches on a query, exposed as API products.
- The most-used search MCP server — hosted at
mcp.exa.ai/mcp, MIT-licensed, with keyless rate-limited access for instant trial.
In an AI-assisted workflow
claude mcp add --transport http exa https://mcp.exa.ai/mcp
# keyless works (rate-limited); add x-api-key from dashboard.exa.ai for real use
# then:
# > Research how teams are handling MCP server auth in production —
# > search broadly, fetch the three best sources, and synthesizeThe MCP toolset is deliberately small after a 2025–26 consolidation: web_search_exa and web_fetch_exa by default, plus an opt-in advanced-search tool with filters. (Older tutorials referencing linkedin_search_exa or deep_researcher_* tools are out of date — those folded into the core tools and the Research API.)
TIP
Exa pairs with Firecrawl as the two halves of agent web-data: Exa finds the right pages; Firecrawl extracts at depth and scale from sites you already know. Plenty of agent stacks run both.
Good to know
Exa Labs raised an $85M Series B (Benchmark, announced September 2025) — the "search engine for AIs" thesis is well-funded and the API surface is moving fast. Pricing is freemium: a monthly free allowance, then metered pay-as-you-go per product tier; enterprise adds zero-data-retention. Like any web-content tool, what it fetches enters your agent's context — treat retrieved pages as untrusted input in injection-sensitive setups.
Frequently asked questions
- What makes Exa different from a normal search API?
- It's built for machines doing retrieval, not people browsing: neural/semantic search over the web with relevance tuned for LLM consumption, a Contents API that returns clean text, highlights, and summaries instead of links to render, structured outputs via output schemas, and deep-search profiles that trade latency for quality when an agent is researching rather than skimming.
- How do I add Exa to Claude Code?
- One verified command: claude mcp add --transport http exa https://mcp.exa.ai/mcp. It even works keyless with rate limits; add an API key from dashboard.exa.ai (x-api-key header) to lift them. The server exposes web_search_exa, web_fetch_exa, and an opt-in advanced search tool.
- Is Exa free?
- Freemium: a monthly free request allowance, then pay-as-you-go metered per thousand requests by product (search, deep search, contents), with enterprise plans adding volume pricing and zero-data-retention. The MCP server itself is MIT-licensed.
Related
- The Best MCP Servers in 2026The MCP servers actually worth connecting in 2026 — Context7, GitHub, Chrome DevTools, Playwright, Serena, Exa, Firecrawl, and the best official vendor servers, by use case.
- Adding MCP Servers to Claude Code: Local, Remote, and Project-ScopedThe complete claude mcp add reference — stdio vs HTTP transports, local/project/user scopes, .mcp.json with env expansion, OAuth via /mcp, and the gotchas.
- FirecrawlThe API to search, scrape, and crawl the web for AI — clean Markdown out of any site, LLM-powered extraction, and a first-class MCP server.
- Data ScientistUse this agent for data analysis — exploration, statistics, SQL, and clear findings. Examples — analyzing a dataset, writing an analytical SQL query, summarizing experiment results.
- How RAG Actually Works: Ingestion, Chunking, Retrieval & RerankingA clear, practical walkthrough of the retrieval-augmented generation pipeline — what each stage does, where it fails, and how the pieces fit together.
- Agent Tool Integration EngineerUse this agent to wire tools and function-calling into an agent loop reliably — clean tool schemas, errors fed back as observations, retries with limits, idempotency, and parallel calls. Examples — "connect our APIs as agent tools", "our agent calls tools wrong / ignores tool errors", "add function-calling with proper error recovery to our agent".
- Web Research PipelineRun a structured web-research pass on a question: plan the searches, find sources via search APIs, fetch and read the best ones, cross-check claims, and synthesize a cited answer — with source quality and disagreements surfaced honestly. Use for 'research X and tell me what's actually true' tasks that need more than one search and less than a day.
- Getting Web Data into AI Agents: Search & Scraping APIs ComparedThe agent web-data layer — Exa for semantic search, Firecrawl for extraction at scale, Tavily for all-in-one, Jina Reader for zero-setup — and how they compose.
- Jina ReaderPrepend r.jina.ai/ to any URL and get LLM-ready markdown — JS rendering, PDFs and Office docs, image captioning, and s.jina.ai for read-the-results search.
- TavilyThe web-access layer for agents — Search, Extract, Crawl, Map, and Research APIs purpose-built for LLMs, behind one key, with a hosted MCP server.