Linkup is a utility for the AI agent stack, providing the real-time data access necessary for agents to perform autonomous tasks. Without a reliable search layer, agents are limited to their training data, making them ineffective for roles in sales, finance, or news monitoring. Linkup connects to this ecosystem by offering an API that fits into agentic loops, with specialized support for frameworks like LangChain and the Vercel AI SDK.

The company is a proponent of the Model Context Protocol (MCP), which simplifies how agents interface with external tools. By providing an MCP server, Linkup makes its search engine a plug-and-play component for any agentic architecture. This matters to builders who need to scale agents in production environments where sub-second latency and factual grounding are mandatory for maintaining user trust.

Grounding LLMs in the live web

Linkup addresses the "airplane mode" limitation inherent in large language models. While LLMs are proficient at reasoning, they are restricted by knowledge cutoffs and lack access to real-time information. Linkup provides a search and retrieval layer that connects these models to the live internet, enabling them to ground their outputs in current, cited data. The company focuses on the "production-grade" segment of the market, where reliability, speed, and accuracy are the primary metrics for success rather than simple hobbyist experimentation.

Three modes of retrieval

The platform is built around three distinct capabilities: Fetch, Search, and Research. Fetch is a utility for retrieving and cleaning the content of specific web pages, removing noise to present a clean input for models. Search provides sub-second results with full-text snippets and citations, intended for high-speed agentic loops. Research is a deeper exploration tool for complex queries that require multi-source synthesis. By separating these functions, Linkup allows developers to optimize for their specific latency and cost requirements.

Linkup has prioritized accuracy benchmarks to prove its utility. The company claims a high F-score on Verified SimpleQA, a benchmark for factual accuracy in search-augmented systems. This focus on measurement is intended to move the industry away from anecdotal "vibe-based" testing and toward reproducible metrics. Their open-source evaluation harness allows teams to verify these performance claims against their own data.

Privacy as a procurement requirement

For enterprise users, the primary barrier to using external search APIs is data privacy. A standard search query sent to a public engine might contain trade secrets or sensitive customer data. Linkup addresses this with a security-first architecture. They provide Zero Data Retention (ZDR) by default, meaning queries are not stored after processing. For highly regulated clients, such as SNCF or global financial institutions, Linkup offers a Bring Your Own Cloud (BYOC) deployment. This allows the entire search indexing layer to run inside the customer’s own infrastructure, ensuring data never crosses an external boundary.

From Paris to the agentic rails

Founded in 2024 in Paris by Philippe Mizrahi, Boris Toledano, and Denis Charrier, Linkup emerged just as the industry shifted from simple chatbots to autonomous agents. Mizrahi, the CEO, has spoken about the need for a legal and sustainable connection between AI developers and content publishers. The company raised a $10M seed round in late 2024 led by Gradient to expand this vision. Linkup now supports a range of enterprise customers, including McKinsey, KPMG, and Artisan AI, who use the API to power agents for sales enablement, legal research, and corporate intelligence. They are currently active in the emerging Model Context Protocol (MCP) ecosystem, providing a standardized way for agents to access their search capabilities.