💼VentureBeat•2026年2月24日較早收集於 14m

Nimble 推出 99% 準確率代理搜尋平台

Post LinkedIn

💼閱讀原文： VentureBeat

#agentic-search #multi-agent #web-scraping #fundingnimble-agentic-search-platformnimble openai anthropic meta

💡99% accurate agentic web search for enterprises—$47M funded multi-agent platform beats LLM guesswork.

⚡ 30-Second TL;DR

有什麼變化

推出時獲 4700 萬美元 B 輪融資，總融資達 7500 萬美元

為什麼重要

Nimble 的平台可彌補代理 AI 的可靠性差距，讓企業應用更穩健。憑藉強大融資，它挑戰 Perplexity 等現有業者，並將代理搜尋定位為 AI 代理基礎設施。

下一步行動

Request Nimble's enterprise demo to integrate agentic web search into your AI data pipeline.

誰應關注：Enterprise & Security Teams

關鍵要點

•推出時獲 4700 萬美元 B 輪融資，總融資達 7500 萬美元
•多代理系統包含 5 層：瀏覽、解析、處理、驗證
•使用 OpenAI、Anthropic、Meta 模型控制瀏覽器，達 99% 準確率
•將網路轉換為 AI 工作流程的決策級資料

🧠 深度解析

背景與延伸：來自公開資料，非原文內容。引用 9 個來源。

🔑 增強重點摘要

•Nimble was founded in 2021 as Thhe Data Company Technologies Inc., focusing initially on real-time web search and data platforms for enterprise AI[2].
•The platform includes a no-code workflow builder for configuring browser-based search agents and a Web Tools SDK with APIs for programmatic search, extraction, and crawling[2][3][5].
•Integrates with enterprise data warehouses like Databricks and Snowflake, as well as Microsoft for incorporating real-time web data into analytics and AI workflows[3][6][8].
•Supports use cases including financial due diligence, e-commerce pricing, media strategy optimization, social listening, and research automation with multilingual and JavaScript-rendered data handling[3][4][5].

🛠️ 技術深入

•Uses AI models to control full web browsers for navigating sites, interacting with dynamic elements, handling layout changes, and retrieving live data, rather than relying on APIs or static scraping[2].
•Applies a governed data layer post-scraping for cleaning, deduplication, joining, aggregation, and schema normalization to produce structured tables[2][3][4].
•Web Search Agents (WSAs) understand query intent, support pre-defined modes like 'general' and 'news', and custom focus modes with specific subagents (e.g., amazon_serp, reddit_discover_posts)[5].
•Features residential proxy network for global coverage, automatic adaptation to page layouts/languages/frameworks, and validation through a data-structuring engine[4].
•Supports long-running multistep workflows with agent coordination for gathering, cross-checking, and validating outputs before downstream integration[2].

🔮 前景展望AI analysis grounded in cited sources

Nimble will expand multi-agent research to support frontier models in production browser systems

Funding accelerates research in multi-agent web search and partnerships with AI labs aim to integrate state-of-the-art models into reliable enterprise environments[3][8].

Enterprises will increasingly rely on Nimble for real-time web data in AI agents

Platform addresses foundational needs for trusted external data in production AI workflows, as validated by integrations with Microsoft and Databricks[3][8].