Tavily pricing and review

Reachkit, which publishes this page, makes cold email software with its own lead database, so it overlaps with Tavily only where Tavily is used to research sales accounts. Every price below was checked on October 15, 2026 against Tavily’s pricing page, docs, help center and terms.

Tavily pricing starts free with 1,000 API credits a month, then runs from $30 to $500 a month on paid plans, or $0.008 per credit on pay-as-you-go. A basic search uses 1 credit. Prices are monthly in USD, checked October 15, 2026.

Tavily’s pricing page shows the free plan, Project and a slider for bigger tiers; the Bootstrap, Startup and Growth prices are from its credits docs and help center. We found no annual billing. Enterprise goes through sales, and the one public figure is a $49,000 12-month list price on AWS Marketplace.

What a credit buys. Advanced search costs 2 credits. Extract costs 1 credit per 5 successful URLs (2 on advanced), Map 1 per 10 pages, and Crawl is the two added together. Failed extracts and maps are free. A Research request takes 4 to 110 credits on the mini model and 15 to 250 on pro, depending on the request.

Billing rules. Credits reset on the 1st of each month whatever your billing date; Tavily doesn’t say whether unused ones carry over. Past your plan, pay-as-you-go charges $0.008 a credit, billed for every $40 of usage, and you can set a monthly limit in billing settings.

What a typical setup costs. Researching 2,000 accounts a month with two advanced searches and one advanced extract of up to 5 URLs each takes 12,000 credits: $94 on Project plus pay-as-you-go, or $100 on Bootstrap. A thousand basic searches a day (about 30,000 credits) fits Startup at $220, against $240 on pay-as-you-go alone.

Tavily lets an AI agent search the live web and read pages through one API. You send a query or URLs and get back ranked results with content snippets, or cleaned page text, ready for a model.

• Search: basic, advanced, fast and ultra-fast depths, with domain, language and exact-match filters
• Extract: up to 20 URLs a call, with an optional query that reranks the returned chunks
• Crawl and Map: walk or list the pages of a site
• Research: an agent that runs multi-step web research and writes up what it found

You connect through a REST API, a Python SDK, LangChain or Vercel AI SDK packages, or Tavily’s official MCP server, a standard way for AI assistants like Claude and ChatGPT to use other apps. Search and Extract also work keyless, with no account, at limits Tavily doesn’t publish.

For sales teams the usual job is account research; a case study on Tavily’s homepage says Rox cut it “from a Week to Seconds”. Since February 2026 Tavily has belonged to Nebius (NASDAQ: NBIS), which says the brand and API continue.

Strengths

• Cheap to try. 1,000 credits every month with no card, keyless Search and Extract, and free access for students (pricing).
• Quick to wire in. G2 reviewers wrote “Integration was effectively a URL swap” (May 2026) and “Easy to integrate” (July 2026), as shown on the AWS Marketplace listing.
• Accurate company lookups. A G2 reviewer testing company enrichment wrote “all ten came back correct”. Tavily also claims “#1 in Company Search Benchmark”, on a benchmark it built.
• Agent access built in. The MCP server works with OAuth, an API key or no key at all.

Drawbacks

• Cost grows with volume. Some third-party reviews put its per-query cost above raw search-result APIs, and one G2 reviewer found it “a bit expensive”; another called the pricing “super friendly”.
• Research is hard to budget. One request can take 4 to 250 credits.
• Some dead links. A GitHub thread reports results pointing to 404 pages of shut-down sites, and three replies say the poster isn’t alone in seeing it.
• Mixed signals in the terms. Fees are non-refundable and plans auto-renew. The terms ask for cancellation 24 hours before billing in one section and 30 days’ notice (where a plan sets none) in another.

Few users have written reviews of Tavily, though its homepage says it’s “Trusted by 3M+ developers”. G2 shows 4.5 out of 5 from 3 reviews, read from search results and the AWS Marketplace mirror, and G2 says that’s too few for buying insight. We found no Capterra listing, and Tavily’s Trustpilot page has no reviews, checked October 15, 2026.

The praise is about fast setup and correct company data; the complaints are about cost at volume and stale links. Much of what ranks for “tavily review” is written by competitors such as Firecrawl, Exa and Apify.

Tavily fits developers and AI teams who need live web results inside an agent and would rather pay per call than run scrapers, including GTM engineers whose agents research accounts: what a company sells, who it serves, what changed lately.

Look elsewhere if nobody on your team writes code, or if what you’re missing is the people to email; Tavily isn’t sold as a contact database. In cold email, research like this is where AI earns its place: qualify accounts with it, then pull named contacts from a lead database. More in lead scraping for cold email.

Exa and Linkup also come up in Tavily comparisons; we haven’t checked their prices.

• Firecrawl: a scrape-first web data API with an open-source core, 1,000 free credits a month and Hobby at $19. See the Firecrawl profile.
• SerpApi: Google and other search results as structured data, free for 250 searches, then $25 a month. See the SerpApi profile.
• Clay: enrichment tables that run many data providers per row, free plan, Launch at $185 a month. See the Clay profile.
• People Data Labs: a B2B data API for person and company records, from $98 a month. See the People Data Labs profile.
• Reachkit (our product): if your Tavily research ends in a cold email list, Reachkit’s Lead Finder finds verified contacts at the companies you name (500M+ people, 1 credit per lead), and Claude or ChatGPT can run that search and build the campaign through its MCP server. It doesn’t search the open web; bring research columns in by CSV as merge tags. From $39/month.