Inside H Company's Computer-Use Agents: Web Search at Production Scale with Linkup

At a Glance

Company. H Company builds a platform to deploy state-of-the-art computer-use agents — agents that don't just browse the web, but act on it, clicking, scrolling, typing, as a human would do.

The Problem. To accomplish their tasks effectively, those agents needed web search capabilities that were fast, high-concurrency, and compliant. Their previous setup couldn't deliver on all three at production scale.

Implementation. Linkup's search API embedded as the retrieval layer across H Company's agent stack, called mid-task wherever an agent needs live web context.

Key Results

• ~1.5s median latency, 3x faster than the previous setup
• Full EU data residency for regulated clients, with no exceptions
• Public API launched with Linkup integrated stack-wide

When your agents don't just browse the web but act on it, search can't be an afterthought. That's the problem H Company had to solve as it scaled toward the public launch of its agentic platform. Its agents needed fast, reliable access to real-time web context.

The Challenge

H Company's agents run at serious scale. At peak, hundreds of agents execute simultaneously, each spawning sub-agents that pull live web context mid-task. That workload puts extreme pressure on any search API:

• High concurrency — bursts of 100+ QPS the moment tasks kick off
• Low latency — every second of delay compounds across parallel agents
• Data sovereignty — regulated clients in finance and other industries require a full EU data residency stack, with no exceptions

Why Linkup

For H Company, the decision came down to what their agents actually needed in production.

Speed. ~1.5s median latency, 3x faster than their previous setup — critical when every additional second of delay compounds across hundreds of parallel agents running at once.

EU-native infrastructure. True European data residency built into the platform, covering H Company's regulated clients with no exceptions.

Grounded results. A search call an agent makes mid-task has to return what's actually on the page, not a plausible-sounding guess at what the web probably says. H Company couldn't afford results that look cited but aren't, especially operating in regulated and demanding industries where auditability is crucial.

The Use Case

Linkup functions as a retrieval layer that H Company's agents call mid-task to stay grounded in real-world context. Key workflows include legacy system lookups, competitive intelligence, web scraping and live data retrieval.

The Result

H Company launched its public Computer-Use Agents API with Linkup's search API integrated across the agent stack. That gives them confidence that EU-regulated clients are fully covered on data residency, along with the headroom to scale as volumes grow.

For platforms where hundreds of agents act on the web in parallel, search infrastructure is either invisible or it's the bottleneck. Linkup lets H Company treat it as invisible.



Ready to give your agents the context they need? Get started for free here. Or, learn more from our team here.