Tavily Raises $25M to Securely Connect Enterprise AI Agents to Web
Across a multitude of industries, companies are rapidly deploying artificial intelligence agents to automate an extensive array of internal tasks. In the financial sector, for instance, these AI agents are proving indispensable for fraud detection, capable of analyzing vast quantities of transaction data in real time. Similarly, sales organizations are leveraging AI agents to meticulously gather intelligence on potential customers, with these digital assistants adept at scouring the web and social media for critical information.
However, for these AI agents to operate effectively and reliably within an enterprise environment, they require secure and compliant access to the internet. They must not only find information from relevant sources but also strictly adhere to company policies, mirroring the diligent approach of a human researcher. The challenge arises when an AI agent is connected directly to a large language model like ChatGPT without company-specific safeguards, a scenario that can lead to highly inappropriate or non-compliant results.
George Mathew, managing director at Insight Partners, articulated this critical concern to TechCrunch, emphasizing the paramount importance of “governance, risk and compliance at the enterprise.” He warned that without proper controls, the widespread adoption of AI agents could quickly devolve into a “wild, wild west.”
It is precisely this challenge that Tavily, a one-year-old startup, aims to address. Insight Partners recently led a $20 million Series A funding round in Tavily, bringing the company’s total funding to $25 million. Tavily’s core offering is a solution that connects AI agents to the web in a manner that is fully compliant with a company’s specific policies and regulatory requirements.
Tavily’s journey began last year as an open-source project called GPT Researcher, created in 2023 by data scientist Rotem Weiss. This consumer-focused project gained significant traction, accumulating nearly 20,000 GitHub stars, primarily because it offered real-time web data access before prominent large language models like ChatGPT integrated their own web search capabilities. After ChatGPT and other LLMs introduced web search, Weiss pivoted his focus, launching Tavily specifically for enterprise clients.
Unlike its open-source predecessor, Tavily provides a comprehensive suite of tools tailored for corporate use. These tools enable AI agents from clients such as Groq, Cohere, MongoDB, and Writer to efficiently search, crawl, and extract structured insights from both public and private data sources. This controlled access ensures that AI agents operate within defined parameters, mitigating risks associated with uncontrolled information retrieval.
While the majority of AI agents in use today are not yet connected to the internet, Tavily harbors ambitious plans. Rotem Weiss stated that the company’s ultimate goal is to facilitate the onboarding of the next billion AI agents to the web. The nascent market for AI agent search tools is, however, attracting competition. Tavily competes with companies like Exa, which secured a $17 million Series A last year from investors including Lightspeed, Nvidia, and YC. Smaller startups such as Firecrawl also offer web search connectivity layers. Furthermore, established players like OpenAI and Perplexity are providing search solutions, though these are primarily geared towards independent developers rather than enterprise-level compliance.
As AI agents become increasingly integral to business operations, the demand for secure, compliant, and efficient web connectivity will only grow, positioning companies like Tavily at the forefront of this critical technological evolution.