Linkup

Linkup develops a web search and retrieval API designed for AI systems rather than human users. Its core proposition is that AI models have knowledge frozen at the end of training, while conventional search engines are optimized for human browsing behavior and return links rather than extracted information. Linkup rebuilt web search for machine consumption, returning sourced, cited answers with full-text snippets that AI applications can process directly.

The platform exposes five endpoints: Search (query to ranked URLs, a sourced answer, or structured JSON, with latency from under one second to about 30 seconds), Fetch (URL to clean LLM-ready markdown or typed JSON, roughly one second), Research (query to a detailed report or precise answer with inline citations, asynchronous, two to twenty minutes), Tasks (batches of up to 100 Search/Fetch/Research calls) and Extract (seed URL plus query to downloadable structured rows). Research is in beta and Extract is in closed beta, granted by request. Customers can tune the index, select which sources and domains to trust, set freshness, and blend in proprietary data. Deployment options include a dedicated access-controlled Private Index and Bring Your Own Cloud, in which the full indexing layer runs inside the customer's own infrastructure or VPC so queries do not leave the environment. Compliance and security features cited include SOC 2 Type II, Zero Data Retention, GDPR and Single Sign-On on every tier, alongside an SLA-backed 99.9% uptime commitment.

The company positions itself as an infrastructure layer for the emerging "web of AI agents," and integrates with the OpenAI SDK wrapper, LangChain, CrewAI and other agent frameworks, with client libraries for Python, Node.js and direct HTTP access.

Founding story

The founding team combined complementary backgrounds in search and enterprise technology: CTO Denis Charrier had previously built one of the first neural music search engines, which was acquired by Spotify; CEO Philippe Mizrahi had built graph-based data search systems at Lyft; and COO Boris Toledano came from McKinsey, where he led digital transformation work in enterprise environments. Observing that AI systems' knowledge is frozen at the end of training and that human-oriented search engines return links rather than usable information, they set out to rebuild web search natively for how AI systems process information.

Business model

Linkup sells API access to its web search, fetch, research, extraction and batch-task endpoints, with a freemium entry point and flexible paid pricing tiers. Beyond metered API usage, it offers enterprise deployment modes — a dedicated Private Index and Bring Your Own Cloud installation of the indexing layer — together with SLA, security and compliance commitments aimed at larger contracts.

Usage-based API access under a freemium model, with paid and enterprise tiers that add private indexes, in-customer-cloud deployment, security certifications and SLA-backed uptime.

Traction

Linkup states it launched its API in 2024 and has since reached thousands of users globally and hundreds of paying customers, ranging from AI startups such as Artisan to large enterprises including KPMG. Other named users include Legora, Polymarket, Cohere, McKinsey, SNCF and H Company. Published customer outcomes cite a 75% increase in adoption rate and a 30% reduction in hallucinations.

Latest developments

In early 2026 Linkup announced a $10M seed round led by Gradient with participation from Elaia, Leblon Capital, Weekend Fund, Seedcamp, Axeleo Capital, Motier Ventures, OPRTRS CLUB and angel investors including founders from Datadog, Mistral, Deel and Dataiku, stating that proceeds would fund technical infrastructure, global team expansion and product development. Alongside the round it launched Linkup /fast, a sub-second web search API. Its Research endpoint is in beta and an Extract endpoint is in closed beta available on request.