Internet Search has already been disrupted by AI — simply check out how readily Google is presenting customers with AI Overviews (summaries of search outcomes) on the high of their outcomes pages, how Bing early on built-in OpenAI's GPT fashions, and the way Perplexity continues to construct by itself AI-driven internet search platform and browsers.
Nimble introduced the launch of its Agentic Search Platform, a system designed to remodel the general public internet into trusted, decision-grade information for AI programs and enterprise workflows.
The launch is supported by $47 million in Collection B financing led by Norwest, with participation from Databricks Ventures and others, bringing the corporate's whole funding to $75 million.
The initiative addresses a basic bottleneck within the present AI period: whereas massive language fashions (LLMs) have gotten extra refined, they typically motive over incomplete or unverifiable exterior data. Nimble’s platform goals to get rid of this "guesswork hole" by offering a ruled information layer that searches, navigates, and validates stay web information in actual time.
In an unique interview with VentureBeat, Nimble co-founder and CEO Uri Knorovich mirrored on the early skepticism relating to his imaginative and prescient of a machine-centric web.
"At any time when we began this firm, and the primary time I went to traders, I advised them the net is constructed for people, however machines are going to be the primary residents of the net," Knorovich recalled. He famous that whereas preliminary reactions labeled him as "too visionary," the present actuality of AI adoption has validated his thesis.
Know-how: Coordinated multi-agent structure
The core of Nimble’s answer is a proprietary distributed structure that orchestrates specialised brokers to carry out duties historically dealt with by human researchers or brittle internet scrapers. In accordance with the corporate's infrastructure documentation, the method is damaged down into 5 distinct layers:
Headless browser and searching brokers: These layers handle the preliminary interplay with a goal area, navigating complicated web site constructions as a human would.
Parsing brokers: These brokers interpret the web page content material, figuring out related information parts throughout varied codecs.
Information processing brokers: This layer aggregates, filters, and cleans noisy web information to provide particular, structured solutions.
Validation brokers: The ultimate step entails verifying the outcomes to make sure accuracy and completeness earlier than supply.
In contrast to customary search engines like google designed for client link-clicking, this structure makes use of multimodal and reasoning capabilities from frontier fashions—together with these from OpenAI, Anthropic, and Meta—to regulate actual browsers. This permits Nimble to navigate dynamic layouts and cross-check outcomes, producing auditable information outputs reasonably than easy textual content summaries.
A brand new paradigm: 'The net is constructed for people, however machines are the primary residents'
Knorovich factors out that the dimensions of AI interplay with the net is basically totally different from human habits. "We, as people, seek for possibly three or 5 choices earlier than we making selections… however day-after-day, Nimble carry out greater than 3.2 million interactions within the internet," he defined. This sheer quantity of billions of month-to-month searches represents a programmatic shift that requires a brand new kind of infrastructure.
The bottleneck for enterprises as we speak, based on Knorovich, isn't the intelligence of the fashions, however the high quality of the info they will entry. "Brokers are the headlines, and correct and dependable internet search is the bottleneck," he said.
Nimble vs. client search: Precision over pace
Knorovich explicitly differentiates Nimble from general-purpose instruments like Google or client AI search assistants.
Whereas Google has constructed a search expertise for shoppers that’s optimized for pace and discovering an area restaurant, enterprises require high-scale, high-accuracy outcomes to make multi-million greenback selections.
"Normal goal internet search software are nice to have a common solutions, equivalent to who’s the spouse Leo lacking," Knorovich remarked through the interview. "However enterprises want deep, granular information, and they should have the flexibility to regulate the search filters, to regulate the regulation, to regulate what’s a trusted supply". In contrast to client AI modes that will summarize a Reddit publish or high-level information, Nimble offers "street-level" data that may be saved immediately in an enterprise system of report.
Product: Bridging the no-code and developer divide
The Agentic Search Platform is delivered via two major interfaces designed for enterprise scalability:
Internet search brokers: A no-code AI workflow builder that allows enterprise groups to explain the info they want and obtain structured information streams with out writing a line of code.
Web instruments SDK: A set of APIs for builders to go looking, extract, and crawl the net immediately from their code. This contains specialised instruments just like the /crawl API for mapping whole domains and the /map API for creating area timber.
The platform is constructed to ship information with better than 99% accuracy — which means fewer than 1% inaccurate or hallucinated information for the whole contents of every search outcome returned — and a latency of 1-2 milliseconds per request.
It integrates natively with main information environments, permitting customers to stream clear information immediately into Databricks, Snowflake, S3, or Microsoft Material.
In the course of the interview, Knorovich emphasised that Nimble is designed to be model-agnostic, working seamlessly with state-of-the-art fashions from OpenAI, Anthropic, and Google's Gemini. This flexibility permits firms to make use of Nimble alongside their present tech stack, whether or not they’re operating fashions within the cloud or on-premise for high-security environments like healthcare or banking.
Case research: Accuracy in motion
Knorovich offered a number of real-world examples of how this "street-level" information impacts skilled workflows. As an illustration, an actual property dealer trying to broaden into a brand new territory doesn't want a high-level abstract from a general-purpose AI.
"If you wish to know what's occurring within the industrial actual property in Atlanta… you're not in search of search that's optimized for the millisecond," Knorovich defined. "You're in search of street-level, neighborhood-level data… information that you may truly see on a desk or obtain to Excel".
One other use case entails main monetary establishments using Nimble for "know your buyer" (KYC) processes. By deploying an autonomous search agent, banks can cross-reference a number of public stories, prison information, and deal with verifications to construct a whole profile of a consumer earlier than they even enter the constructing. The objective, Knorovich famous, is to supply the "exterior fact" that exists outdoors a corporation's inner firewalls.
Enterprise licensing and compliance
Nimble differentiates itself from legacy scraping instruments via a rigorous deal with governance and belief. The platform is "compliant-by-design," holding certifications for SOC2 Kind II, GDPR, CCPA, and HIPAA.
Pricing is structured to help each experimental startups and high-scale enterprise operations, aligned with the quantity and depth of knowledge retrieved.
"Pricing needs to be aligned with the worth that the consumer is getting… due to this fact, we’re pricing by the quantity of searches that you simply're operating," Knorovich mentioned.
Search and reply APIs: Commonplace search inputs value $1 per 1,000, whereas the "Reply" operate—which offers reasoning primarily based on search outcomes—prices $4 per 1,000.
Managed companies: For bigger organizations, managed tiers begin at $2,000 monthly (Startup) and scale to $15,000 monthly (Skilled) for limitless brokers and precedence help.
Proxy entry: A community of over 1 million residential proxies is offered beginning at $7.50 per GB
Neighborhood and consumer reactions
The transition to agentic search has already been operationalized by a number of Fortune 500 firms and AI-native startups:
Julie Averill, former CIO at Lululemon, said that pricing intelligence which as soon as took weeks to evaluate can now be responded to in minutes by placing management within the fingers of an agent.
Itamar Fridman, CEO and Co-founder of Qodo, famous that the platform’s scalability was "essential in growing extra strong and dependable AI programs" by feeding LLMs with high-quality information.
Dennis Irorere, Information Engineer at TripAdvisor, highlighted that the platform simplifies the extraction of structured information from complicated sources, which he described as "transformative" for his function.
Grips Intelligence reported scaling to over 45,000 e-commerce websites utilizing Nimble’s Internet API to ship real-time pricing and product information.
Alta makes use of the platform to energy hundreds of thousands of AI-driven go-to-market workflows day by day, reporting 3–4× deeper context and >99% reliability
Collection B to speed up multi-agent internet search and information governance
The $47 million Collection B funding introduced alongside the platform can be used to speed up analysis in multi-agent internet search and additional develop the ruled information layer.
The spherical noticed participation from a large ecosystem of traders, together with Goal World, Sq. Peg, Hetz Ventures, Sluggish Ventures, R-Squared Ventures, J-Ventures, and InvestInData.
Andrew Ferguson, VP of Databricks Ventures, famous that Nimble enhances their Information Intelligence Platform by offering a "real-time internet information layer" that extends workflows past inner sources. This strategic funding indicators a shift within the business towards prioritizing "exterior fact" to floor mission-critical AI functions.
For Knorovich, the way forward for the net belongs to programmatic interplay. "Programmatic internet search is the place we’re constructing in direction of," he concluded. By transferring away from legacy information distributors and brittle scrapers, Nimble goals to supply the real-time construction wanted for AI to behave with confidence in the true world.
