Tavily Review, Pricing & Alternatives (2026)

Tavily is a hosted web search and content service operated by AlphaAI Technologies Inc. for developers building LLM applications and autonomous agents. Its documentation describes APIs that search, scrape, filter, and extract material from online sources. The company says its developer community exceeds one million users and its SDKs receive more than three million monthly downloads. Nebius announced an agreement to acquire Tavily in February 2026, and a later filing records $177.3 million in cash paid at closing, with additional possible payments tied to employees and performance targets. Tavily continues under its own brand, while its team and technology now sit within Nebius’s AI cloud business. The company does not publish a founding year or total funding figure in the cited materials.

Tavily’s main workflow separates discovery from deeper site operations. Search returns current web results, with basic requests costing one credit and advanced requests costing two. Extract processes known URLs, Map discovers pages across a site, and Crawl combines mapping with extraction. Research performs multistep searches, source synthesis, and report generation with citations. Developers can call these functions through HTTP APIs, Python and JavaScript packages, the tvly command-line tool, or a remote MCP server. The CLI supports JSON output for scripts and agent pipelines. A typical RAG application sends a question to Search, passes the returned content to a language model, then calls Extract or Crawl when the initial result set lacks enough detail. Domain filters, parsing controls, response limits, and search depth shape the request.

Tavily is a retrieval and content-processing service rather than a published general-purpose foundation model. Its proprietary systems score, filter, and rank sources, while Research offers mini and pro modes with different depth and credit ranges. The documentation says Search can aggregate up to 20 sites and lets developers control search depth, domains, parsing, context, and response tokens. Recent documented additions include Crawl, Map, Research, the remote MCP server, Agent Skills, and CLI workflows. Tavily recommends advanced search and multiple chunks per source when applications need stronger supporting material, although those settings consume more credits. The company does not publish the underlying ranking model or offer an open-source index. Its product documentation centers on retrieval APIs and content operations rather than model training or model hosting.

Tavily’s free Researcher plan includes 1,000 credits each month, with no credit card required. Project costs $30 for 4,000 credits, Bootstrap costs $100 for 15,000, Startup costs $220 for 38,000, and Growth costs $500 for 100,000. Pay as you go costs $0.008 per credit, while Enterprise pricing is custom. Basic Extract charges one credit per five successful URLs, advanced Extract charges two, regular Map charges one credit per ten successful pages, and Map with instructions charges two. Research ranges from 4 to 110 credits for mini and 15 to 250 credits for pro. Crawl combines mapping and extraction charges, so larger site jobs consume credits faster than Search. Project, Bootstrap, Startup, and Growth plans reduce the effective credit rate compared with standalone usage.

Tavily’s main limitation is cost predictability for deep, multistep work. Research can consume up to 250 credits per pro request, and Crawl combines two metered operations. The hosted service does not provide a self-hosted index, and its privacy policy says portions of query data may be used to improve responses unless a contract says otherwise or the account opts out. Exa offers search modes, contents, answers, monitors, and deep-search products. Perplexity focuses on generated answers with citations, while Brave Search API provides conventional web search access. Tavily is the stronger fit when one API must cover retrieval, extraction, site mapping, crawling, MCP, and cited research. Choose it for agent pipelines that need current web material and programmable site operations.