Web Search by Exa

• Semantic web search — You need to find pages by meaning, not exact keywords (e.g. "blog posts about using vector databases for recommendations").
• Targeted research — You need research papers, news, tweets, company intelligence, or professional profiles, with category-level filtering.
• Content extraction — You have a specific URL and want clean, structured content from it, including JavaScript-rendered pages or PDFs.
• Deep research reports — You need a multi-step research agent to explore a topic, synthesize findings, and return a detailed report with grounded citations.
• Agentic pipelines — You're building a multi-step agent workflow that requires fresh, token-efficient web context.

• Local file operations — This skill is web-only; it cannot read or process files on your local machine.
• Non-web tasks — Any task that doesn't involve retrieving content from the internet is outside this skill's scope.
• Offline or air-gapped environments — All tools require outbound HTTPS access to mcp.exa.ai.

Exa's web_search_exa finds pages by meaning, not keyword matching. Write natural-language descriptions of the page you want and Exa returns the most semantically relevant results, with configurable result counts, freshness, and content size.

web_search_advanced_exa adds domain allow/block lists, date range filters, content category targeting (research papers, news, tweets, people, companies, financial reports), freshness controls via maxAgeHours, and multiple content extraction modes including highlights, full text, and LLM-generated summaries with optional JSON schema output.

company_research_exa delivers one-call company overviews including recent news, funding, and competitors. people_search_exa finds LinkedIn profiles and professional bios by role, location, or expertise.

get_code_context_exa searches GitHub repositories, Stack Overflow, and official documentation to surface relevant code examples and API usage patterns for any library or framework.

crawling_exa extracts clean markdown content from any specific URL, handling JavaScript-rendered pages, PDFs, and complex layouts — no scraping setup required.

deep_researcher_start / deep_researcher_check run an async multi-step research agent that searches, reads, and compiles a detailed report. deep_search_exa provides a single-call alternative returning a synthesized answer with grounded citations, and supports structured JSON output via outputSchema.

Use web_search_advanced_exa with category: "research paper" to find arXiv papers and peer-reviewed articles on a topic, then crawling_exa to extract full content from specific papers. Combine with deep_researcher_start to compile a structured literature review.

Run company_research_exa to get an overview, news, and competitor landscape for any company in a single call. Follow up with web_search_advanced_exa filtered by domain or date range to track recent funding announcements or product launches.

Use people_search_exa or web_search_advanced_exa with category: "people" to find professionals by role, company, location, and skill set. Pass a known LinkedIn profile URL as the query to web_search_advanced_exa to surface semantically similar candidate profiles.

Use get_code_context_exa inside an agentic coding workflow to retrieve up-to-date implementation examples, API patterns, and documentation from GitHub and Stack Overflow without leaving the agent context.