Tavily
Also known as: Tavily Search, Tavily API
A web access API for AI agents. It offers real time search, extraction, crawling, mapping and cited research, returned as clean content ready for LLMs through an API, SDKs, a CLI and remote MCP.
Tavily is a web access layer for AI agents. A single API gives an agent real-time web search, content extraction from chosen URLs, site mapping and crawling, and a Research endpoint that plans, searches and returns a cited report. Results come back as clean, LLM-ready content rather than raw HTML, with controls for search depth, domains and the number of results.
Developers reach Tavily through its REST API, Python and JavaScript SDKs, a CLI that also installs Agent Skills for coding agents, and a remote MCP server that supports OAuth, and its documentation covers integrations with most agent frameworks and model providers. Enterprise customers can generate and manage API keys programmatically with permission levels, limits and expiry, and paid plans get per-request usage logs, which record each call but not its input or output.
Nebius Group acquired Tavily in February 2026, and Tavily continues to sell under its own brand. Pricing is credit-based: a free plan with 1,000 credits a month, monthly plans from $30 for 4,000 credits up to $500 for 100,000, pay as you go at $0.008 per credit, and custom Enterprise terms.
Vendor details
Canonical URL
https://www.tavily.com
Category
Agent infrastructure
Subcategory
Web search and retrieval
Funding status
Acquired by Nebius Group, which reports acquiring 100% of Tavily on 19 February 2026 (Nebius Q1 2026 filing); Tavily continues to sell under its own brand.
Company status
acquired
Use cases & customers
Primary use cases
Target customers
Deployment options
Integrations
A REST API, Python and JavaScript SDKs, a CLI with Agent Skills and a remote MCP server with OAuth. Integrations cover OpenAI, Anthropic, Google ADK, the Vercel AI SDK, LangChain, LlamaIndex, CrewAI, Mastra, n8n, Zapier and dozens more, with listings on Snowflake, Amazon Bedrock AgentCore, Azure AI Foundry, Databricks and IBM watsonx Orchestrate.
In practice
Your agent answers questions from stale model knowledge and sometimes invents facts. You add a Tavily search call to the loop, so it pulls fresh, ranked web context and citations before the model responds.
You need clean article text from a list of URLs, not messy HTML. Tavily Extract returns clean content in markdown for up to twenty URLs per call, ready for the model to read.
You are on LangChain and want web search without wiring a provider yourself. Tavily ships a LangChain integration and a remote MCP server, so you drop it in and start grounding answers.
Sources & related URLs
Related / legacy domains
Agentic Index coverage score
4.5 / 14 capabilities · 32%
| Integrations & Tool Calling | Partial |
|---|---|
|
An agent gets read access to the public web as tools (search, extract, crawl, map and research), called through Tavily's API, SDKs or a remote MCP server. Tavily takes no actions in other systems, and there are no credentials per user to scope, revoke or rotate. Sourcedocs.tavily.com/llms.txtread 2026-09-22 |
|
| Workflow Orchestration | Not documented |
|
Workflow logic lives in the customer's own framework, along with the customer's agent. Tavily has no workflow runtime of its own in which the customer can sequence, branch, retry or route steps. Its Research endpoint runs a fixed pipeline, planning, searching and synthesizing a cited report, and the customer cannot define, branch or sequence steps in it. Sourcedocs.tavily.com/documentation/api-reference/endpoint/researchread 2026-09-22 |
|
| Knowledge Grounding & RAG | Not documented |
|
Searches and extraction run against the public web. Tavily keeps no retrieval structure over the customer's own documents or data, and has no ingestion, retrieval or RAG layer that grounds agents in company data. Its Crawl to RAG and hybrid research examples show customers building their own knowledge base with other tools. Sourcedocs.tavily.com/llms.txtread 2026-09-22 |
|
| Human Oversight & Guardrails | Not documented |
|
The service retrieves public web content and takes no actions in other systems. Tavily names no approval step, consent checkpoint, review queue, escalation rule, guardrail, pause or resume control, or policy on what an agent does with that content. API keys carry credit limits and expiry for the account. Sourcedocs.tavily.com/llms.txtread 2026-09-22 |
|
| Security, Identity & Governance | Partial |
|
Enterprise customers generate and deactivate API keys programmatically, with admin or user permission levels, scopes, development or production types, credit limits per key and expiry. Organization usage is reported per key, and the remote MCP server supports OAuth 2.0. Tavily holds SOC 2 Type II and ISO 27001, and its Trust Center at trust.tavily.com keeps the reports behind a portal instead of on the public site. Tavily does not say whether it supports SSO. SourceTavily, tavily.com, trust.tavily.com and docs.tavily.com Enterprise API keysread 2026-10-05 |
|
| Observability & Auditability | Partial |
|
On paid plans the Logs API returns usage logs per request for the account's or organization's keys, filterable by date and endpoint, and the Usage and Organization Usage APIs report credits and request counts per key. Logs never include the input or output of a request, so what the agent asked and received cannot be inspected. Sourcedocs.tavily.com/documentation/api-reference/endpoint/logsread 2026-09-22 |
|
| Memory & State Persistence | Not documented |
|
Each request is answered independently, with no agent context kept across calls. Research tasks are retrieved by request ID until they finish. There is no memory layer. Sourcedocs.tavily.com/llms.txtread 2026-09-22 |
|
| Deployment & Data Residency | Not documented |
|
Customers reach Tavily as a hosted API. It is listed on the Snowflake Marketplace as a native app installed into the customer's Snowflake account, and on Amazon Bedrock AgentCore, Azure AI Foundry and Databricks. The Snowflake app is configured with a Tavily API key and calls Tavily's hosted service, so the search itself does not run in the customer's environment. Tavily names no self hosted, private cloud or region option. Sourcedocs.tavily.com/documentation/partnerships/snowflakeread 2026-09-22 |
|
| Prebuilt Agents, Templates & Packs | Partial |
|
The Research endpoint is a prebuilt research agent that plans, searches, analyzes sources and returns a cited report, in mini, pro or auto modes, streamed or polled. Tavily also publishes Agent Skills for coding agents and an Agent Toolkit of reference research agents as code. There is no browsable catalog of prebuilt workflows, templates or agents for specific roles, and the Examples Hub is tutorial code. Sourcedocs.tavily.com/documentation/api-reference/endpoint/researchread 2026-09-22 |
|
| Triggers & Channel Coverage | Not documented |
|
Work starts only when an agent or application calls Tavily, which is a request and response API. It has no schedule, webhook, event subscription or channel that starts an agent. Sourcedocs.tavily.com/llms.txtread 2026-09-22 |
|
| Model Flexibility & Routing | Full |
|
Tavily serves web data to whatever model the customer's agent runs. Integrations cover OpenAI and Anthropic tool calling, Google ADK, the Vercel AI SDK, LangChain, LlamaIndex, CrewAI, Mastra and others, and any MCP client can connect, each with the customer's own model. Only the Research endpoint picks a model tier that Tavily runs (mini, pro or auto). The customer chooses the model for its own agent and brings its own providers. Sourcedocs.tavily.com/llms.txtread 2026-09-22 |
|
| APIs, SDKs & MCP Extensibility | Full |
|
Outside callers reach Tavily through a REST API with an OpenAPI specification covering search, extract, crawl, map, research, usage, logs and Enterprise key management. Python and JavaScript SDKs, a CLI with Agent Skills and a remote MCP server with OAuth 2.0 sit alongside it, and Tavily is embedded in dozens of agent frameworks and platforms. Sourcedocs.tavily.com/documentation/api-reference/introductionread 2026-09-22 |
|
| Testing, Debugging & Optimization | Not documented |
|
A Feedback endpoint (beta) collects feedback on search results to improve Tavily, and the RAG Evaluation example is tutorial code. Tavily's published benchmark results cover its own search. Tavily names no testing, scoring or quality gates for the customer's agent or its workflows. Sourcedocs.tavily.com/llms.txtread 2026-09-22 |
|
| Browser & Computer Use | Partial |
|
Web pages that have no API can be read and traversed through Tavily's Extract, Crawl and Map endpoints, which scrape them and return clean content. Tavily names no interactive control of a real browser, such as clicking, typing, filling forms or operating a logged in session. Sourcedocs.tavily.com/documentation/api-reference/endpoint/crawlread 2026-09-22 |
|
The Agentic Index coverage score grades every vendor Full, Partial or Not documented against the same 14 buyer facing capabilities, from public evidence only. Each capability links to how all vendors in the index score on it. How this evidence is graded
Product updates
Tavily overhauled its search system's core components, including evidence ranking, contradiction handling, and index coverage. These updates allow the search API to better identify semantically redundant or conflicting snippets, improving retrieval accuracy on the SealQA and SimpleQA benchmarks.
Bears on: MCP / tool calling / API
View sourcexAI has launched a native Tavily search plugin within Grok Build, its developer platform. This integration allows developers building on the Grok ecosystem to invoke Tavily's agent-optimized web search and extraction API directly to retrieve structured, LLM-ready results.
Bears on: Integrations
View sourceTavily achieved ISO/IEC 27001:2022 certification, independently validating its security program, infrastructure, and internal controls against recognized information-security standards.
Bears on: Security / enterprise
View sourcePricing
Free 1,000 credits/mo · Project $30/mo (4,000 credits) · pay as you go $0.008/credit · Enterprise custom
Credit system metered by operation type and depth (basic search 1 credit, advanced 2, extract 1 credit/5 URLs, map 1 credit/10 pages); pay as you go or subscription, plus Project and Enterprise tiers
Included quota
Researcher: 1,000 credits a month free. Project: 4,000. Bootstrap: 15,000. Startup: 38,000. Growth: 100,000. Pay as you go beyond a plan at $0.008 per credit.
What is public
Every named plan, its monthly credits and price, the pay-as-you-go rate and per-operation credit costs are public; only Enterprise is custom.
Billing mechanics
Metered credits by operation type and depth, drawn from a monthly subscription allowance or paid per credit on pay as you go. Project and Enterprise tiers raise limits.
Cost watchouts
Advanced depth doubles search cost, the optional answer synthesis and extract calls add credits, and per loop search in busy agents multiplies spend; a heavy research agent can run far above the headline per credit rate.
Variable cost rationale
Cost scales directly with the number and depth of calls. Agents that hit search on every reasoning loop, use advanced depth, or extract many pages accumulate credits quickly. At 100,000 basic searches a month, spend is roughly $800 on pay as you go before subscription discounts.
Additional watchouts
Advanced search depth doubles the search credit cost, and high frequency search loops accumulate credits fast; price your real mix of search, extract, and research calls rather than the cheapest line item.
Overage / add-ons
Usage is metered in credits by operation type and depth; you draw down monthly subscription credits or pay per credit on pay as you go. Exceeding a plan means buying more credits or moving to a higher tier.
Sales call required
No, self serve available
Free / trial
Free tier: 1,000 API credits/month, no credit card. Free for students.
Lowest paid plan
Project, $30 a month for 4,000 credits ($0.0075 per credit)
Commercial notes
Self serve and developer first; Nebius Group acquired Tavily on 19 February 2026 and Tavily continues under its own brand.
Key ambiguities
Enterprise pricing and volume are custom. Credit costs vary by operation and depth (a basic search is one credit, an advanced search two).
Cancellation / refund
Pay as you go and self serve subscriptions have standard cancellation. Enterprise terms are contractual.
Support SLA / resale
Community and self serve support on lower tiers; higher limits and support on Project and Enterprise.
Missing data
Enterprise pricing and volume terms.
Related vendors
- AgentOps — Agent observability and debugging platform: open source SDKs trace…
- AgentPhone — AgentPhone gives the AI agents a team builds their own US or…
- Agno — Python agent framework and AgentOS runtime (formerly Phidata) for…
- AIsa — Resource and payment gateway for AI agents, with one key to 110+…
- Albato — Albato is a no code integration platform that sells an embedded,…
- AlphaBitCore — AI control plane for regulated financial firms: one gateway enforces…
Alternatives to Tavily
The closest documented capability profiles to Tavily among agent infrastructure platforms tracked by Agentic Index, ordered by similarity on the same 14 point evidence the rankings use. No vendor pays for placement.
- Exa6.5 / 14Adds documented Memory & State Persistence and Triggers & Channel CoverageTavily vs Exa →
- AIsa6.0 / 14Adds documented Human Oversight & Guardrails and Deployment & Data Residency
- Jina AI5.0 / 14Adds documented Knowledge Grounding & RAG and Deployment & Data Residency
- Circuit & Chisel4.5 / 14Adds documented Human Oversight & Guardrails and Memory & State Persistence
- Prime Intellect6.5 / 14Adds documented Workflow Orchestration and Deployment & Data Residency, among others
- AgentOps5.0 / 14Adds documented Deployment & Data Residency and Testing, Debugging & Optimization
Similarity is computed from each vendor's Agentic Index coverage score evidence, axis by axis, not from the totals. How this evidence is graded