Conntour raises $7 million for AI video search in security systems

Conntour sells AI search for security camera footage, a seed round bets on queryable surveillance at scale, the marginal cost of watching approaches zero

Conntour, a two‑year‑old startup building an AI “search engine” for security video systems, has raised $7 million in seed funding from General Catalyst, Y Combinator, SV Angel and Liquid 2 Ventures, according to TechCrunch. The company says its software lets security staff type natural‑language queries—“someone in sneakers passing a bag in the lobby”—and retrieve matching clips across recorded footage or live feeds.

The pitch is not new surveillance hardware but indexing: turning video into a database that can be searched, summarized and turned into incident reports. Conntour says it can also run continuous monitoring based on preset rules and push alerts automatically. Its founder and CEO, Matan Goldner, told TechCrunch the round closed within 72 hours after a burst of investor meetings—an anecdote that fits a market where camera networks are already installed and software is the lever that turns them into something closer to a real‑time dossier.

The company’s technical claim is scalability. Goldner says Conntour can monitor up to 50 camera feeds on a single consumer GPU such as Nvidia’s RTX 4090 by routing each query to the cheapest combination of models and logic needed to answer it. That matters because the practical constraint on video analytics has often been compute cost: it is one thing to run object detection on a handful of feeds, another to do it across thousands of cameras in warehouses, transport hubs, retail chains or municipal networks. If the compute per camera falls, the economic barrier to “search everything” falls with it.

That cost curve shifts behaviour. Traditional CCTV is largely reviewed after an incident, because human attention is scarce and expensive. Queryable footage makes retrospective review cheap and routine, and automated alerts make continuous monitoring plausible even for organisations that previously could not staff it. The result is not only more cameras, but more reasons to keep footage longer, to integrate it with access logs and ID systems, and to make internal investigations depend on vendor tools rather than on‑site operators.

TechCrunch places Conntour’s launch against a backdrop of growing scrutiny of how law enforcement accesses private camera networks. The outlet notes controversy around Immigration and Customs Enforcement tapping into Flock’s camera network, and criticism of Ring for features that could make it easier for police to request neighbourhood footage. Conntour says it is “picky” about customers, but also cites “several large government and publicly‑listed customers,” including Singapore’s Central Narcotics Bureau.

Control in this market tends to follow who owns the model, the logs and the update channel. Conntour says its system can be deployed on‑premises, in the cloud, or as a hybrid, and that it plugs into existing security systems. In practice, the more organisations rely on a vendor’s models to interpret footage, the more that vendor becomes the gatekeeper for audit trails, retention defaults, and what can be searched without a bespoke engineering project.

For security departments, the attraction is straightforward: being able to answer “what happened” in minutes instead of hours, and to demonstrate diligence with searchable records. The bill arrives elsewhere—in storage, in integration work, and in the quiet expansion of what is considered normal to look for once the tooling makes it easy.

Conntour’s funding round is small by Silicon Valley standards, but it is aimed at a specific bottleneck: making surveillance footage queryable at scale. The product’s value proposition is that the hardest part of watching is no longer watching—it is typing the right question.