Tavily
by Tavily (Nebius Group)
Real-time web search, extraction and crawling built for AI agents
Tavily is a search, extraction and crawling API purpose-built for AI agents and RAG pipelines. It runs its own index rather than reselling Google rankings, and returns clean, LLM-ready content in a single call, so engineering teams get grounded, citable web data without building and maintaining their own scraping, ranking and parsing stack.
Tavily is a web-access API built specifically for AI agents and retrieval-augmented generation pipelines, replacing the scrape-and-parse plumbing teams would otherwise write themselves. Rather than reselling Google or Bing rankings, it operates its own index and returns cleaned, LLM-ready content in a single call. Five endpoints cover the workflow: Search (fast ranked results), Extract (structured page content), Map (URL discovery across a site), Crawl (whole-site extraction) and Research, an agent that runs multi-step investigations and returns a synthesised report. The vendor publishes a 180 ms p50 latency on Search, a 99.99% uptime SLA, more than 300 million monthly requests and 2 million developers; an independent 2025 comparison of five search APIs measured 93.3% accuracy on SimpleQA and a 1.9-second average response, third-fastest of the group. Official Python and JavaScript SDKs plus first-class LangChain, LlamaIndex and Model Context Protocol integrations make it a common default inside agent frameworks, and the Python SDK is MIT-licensed with roughly 1,370 GitHub stars and active releases through mid-2026. Named users include IBM, MongoDB, LangChain, JetBrains, AWS, Mastercard, Cohere and Groq. Founded in Israel in 2024 by Rotem Weiss, Assaf Elovic and Yuval Rozio, Tavily raised $25 million led by Insight Partners and Alpha Wave Global before Nebius agreed in February 2026 to acquire it for a reported $275 million — as much as $400 million with performance earn-outs — folding agentic search into Nebius's AI cloud platform.
Engineering teams shipping production AI agents or RAG systems that need grounded, current web data and do not want to own a crawling, ranking and content-cleaning stack.
One API call returns ranked, de-duplicated, LLM-ready page content, so agents answer from live sources instead of stale model weights.
At a Glance
- Category
- Agent Development
- Pricing
- Freemium, Usage-based, Subscription, Contact for pricing
- Target Market
- CTOs, Enterprise Developers, AI/ML Engineers, Data Scientists, Platform Engineering Leads
- Deployment
- API-based, Cloud-only
- Founded
- 2024
- Customers
- Vendor claims 2M+ developers; Nebius cited 1M+ developers and 3M monthly SDK downloads at acquisition. Named users include IBM, MongoDB, LangChain, JetBrains, AWS, Mastercard, Cohere and Groq.
Key Features
- ✓Search endpoint
Returns ranked, cleaned results from Tavily's own index with configurable search depth and domain filtering, at a published 180 ms p50 latency.
- ✓Extract endpoint
Pulls structured, LLM-ready content out of specific URLs in batches, and charges nothing for extractions that fail.
- ✓Crawl and Map endpoints
Map discovers URLs across a whole site and Crawl extracts them, so an agent can ingest documentation or a competitor site in one job.
- ✓Research endpoint
An agent that runs multi-step web investigation and returns a synthesised report, priced dynamically at 4-250 credits depending on model tier.
- ✓Framework and MCP integrations
Native LangChain, LlamaIndex and Model Context Protocol support plus Python and JavaScript SDKs, so it drops into existing agent stacks without glue code.
- ✓Enterprise security posture
SOC 2 Type II and ISO 27001 with zero data retention by default, which is what lets regulated teams put it in production paths.
Capabilities
Use Cases
- •Grounded customer-facing assistant
A support or sales assistant queries Tavily before answering so replies cite live product, pricing and policy pages rather than stale training data.
- •Automated deep research
An analyst agent calls the Research endpoint to investigate a market or vendor across many sources and returns a cited brief in minutes.
- •Company and lead enrichment
A GTM pipeline crawls company sites and news to fill firmographic fields, which one AWS Marketplace reviewer cites as their production use.
- •Coding agent with live docs
A coding agent maps and extracts a library's documentation site so generated code matches the current API rather than a year-old snapshot.
- •Continuous competitive monitoring
A scheduled agent crawls competitor pricing and release pages daily and flags material changes to a product or strategy team.
Ideal For
Best For
- ✓Grounding LLM answers in current web sources for RAG pipelines
- ✓Deep-research agents that need multi-step web investigation with citations
- ✓Competitive and market monitoring agents that must read many sites per run
- ✓Enriching CRM or company records from public web pages at scale
- ✓Adding live web access to LangChain, LlamaIndex or MCP-based agent stacks
Not Ideal For
- ✗SEO and rank-tracking teams — Tavily is its own engine and cannot return actual Google or Bing SERP positions, which is what rank monitoring requires
- ✗Entity-heavy retrieval (people, companies, publications) where a neural index such as Exa's specialised indexes reportedly outperforms it on semantic and multi-step queries
- ✗Teams with hard data-residency or air-gap requirements — it is an API-only SaaS with no self-hosted deployment
- ✗Very high-volume scraping on a tight budget, where a raw scraping API costs a fraction per request
Integrations
Deployment
Market & Ratings
Vendor claims 2M+ developers; Nebius cited 1M+ developers and 3M monthly SDK downloads at acquisition. Named users include IBM, MongoDB, LangChain, JetBrains, AWS, Mastercard, Cohere and Groq.
Market Analysis
Pros
- ✓Single call returns ranked and cleaned content, which measurably cuts the glue code an agent needs versus split search/contents APIs
- ✓Genuinely generous free tier (1,000 credits/month, free for students) plus pay-as-you-go, so evaluation costs nothing
- ✓Independently benchmarked at 93.3% accuracy on SimpleQA, strong for RAG grounding
- ✓SOC 2 Type II, ISO 27001 and zero data retention make it approvable inside regulated enterprises
- ✓Ubiquitous framework support means most agent stacks can adopt it in a few lines
Cons
- ✗Cannot reproduce real Google or Bing SERP rankings, so it is the wrong tool for SEO or rank monitoring — an explicit limitation in independent comparisons
- ✗Independent comparisons describe its retrieval as keyword-leaning and weaker than Exa's neural index on semantic and multi-step entity queries
- ✗Credit metering is hard to forecast: the Research endpoint alone ranges 4-250 credits per call, so agent cost per task varies by more than an order of magnitude
- ✗Very thin independent review footprint — 3 reviews on AWS Marketplace, effectively nothing on G2 or Hacker News beyond acquisition coverage
- ✗Now owned by Nebius; a core dependency changing hands introduces roadmap, pricing and packaging uncertainty that did not exist as an independent vendor
Pricing
Researcher
$0
- ✓1,000 API credits per month
- ✓No credit card required
- ✓Email support
- ✓Free for students
Pay As You Go
From $0.008/credit
- ✓No monthly commitment
- ✓Cancel anytime
- ✓Email support
Project
From $30/mo
- ✓4,000 API credits per month at the base tier
- ✓Slider-adjustable credit volume
- ✓Higher rate limits
Enterprise
Contact for pricing
- ✓Custom credit allowance and rate limits
- ✓Enterprise SLAs
- ✓Zero data retention
- ✓SOC 2 Type II and ISO 27001
Metered in API credits, not seats: basic Search costs 1 credit and advanced Search 2, Extract is 1 credit per 5 URLs (2 advanced), Map is 1 credit per 10 pages, and the Research endpoint swings from 4 to 250 credits per call depending on model tier — so agent workloads are genuinely hard to forecast before you instrument them. List pricing is published up to the Project tier (from $30/mo for 4,000 credits) with pay-as-you-go at $0.008/credit; an independent comparison put effective cost at roughly $5-8 per 1,000 searches, cheaper than SerpAPI but far above raw scraping APIs. Enterprise is quote-only, and the AWS Marketplace listing shows a 12-month contract at $49,000 with a 90-day 1,000-credit trial.
Security & Compliance
Connect
Sources
This page was written from 9 sources, 7 on domains other than tavily.com.
- 1.tavily.com — tavily.comvendor
- 2.tavily.com — pricingvendor
- 3.docs.tavily.com — api credits
- 4.nebius.com — nebius announces agreement to acquire tavily to add agentic
- 5.aws.amazon.com — prodview myijjwd7qoky4
- 6.github.com — tavily python
- 7.timesofisrael.com — dutch based nebius buys israeli startup that connects ai age
- 8.dev.to — best serp api comparison 2025 serpapi vs exa vs tavily vs sc
- 9.hn.algolia.com — hn.algolia.com
Stay Ahead of the Curve
Weekly enterprise AI insights for technology leaders. No spam, no vendor pitches—unsubscribe anytime.
SubscribeRelated Products
Sim
Open-source, Apache-2.0 AI workspace for building, deploying and governing AI agent workflows
Zep
Context layer for AI agents: temporal knowledge graphs that give agents governed memory of users, data and documents
Restate
Durable execution runtime that keeps AI agents and backend workflows running through crashes and failures
Mastra
Open-source TypeScript framework and platform for building, deploying and observing production AI agents and workflows