Tabnine

Codota, 2013: Technion research that became a product

The company that later became Tabnine was founded in 2013 in Tel Aviv as Codota, by Dror Weiss and Dr. Eran Yahav, a professor of computer science at the Technion. The original technology was built on more than a decade of academic research in code analysis and code prediction done in Yahav's lab, before the era of large language models as we know them today. The core idea was to analyze huge amounts of existing code and use the patterns found in it to suggest to developers the next line or function they were about to write. Weiss, himself a Technion computer science graduate, served as CEO from day one, alongside Yahav, who stayed active in academia and was also the CTO of the company he co-founded.

The acquisition that brought the name: Canada's TabNine, 2019

In 2019, Codota acquired a much smaller Canadian startup called TabNine, which had developed its own neural-network-based code completion tool. At the time, it was seen as innovative compared with the rule-based approaches common in the market. After the acquisition, Codota gradually adopted TabNine's technology and approach. In May 2021, after launching the first version of its coding assistant based on its own large language model, the company changed its official name from Codota to Tabnine and adopted it as its only brand from then on. This was not just a cosmetic rebrand but a sign of a real shift in the technology itself: from a tool based mainly on static code analysis to a tool based on generative code creation with its own large language model.

Privacy as the main difference from GitHub Copilot

Over the years, Tabnine built a position different from most of the big competitors in the coding assistant market, led by GitHub Copilot: a stated "no-train-no-retain" policy. This means not training on customers' code and not storing it beyond the time needed to process it. Alongside this, the company offers several deployment modes, all at the same price: public cloud, virtual private cloud (VPC), the organization's own on-premises infrastructure, and even a fully isolated (air-gapped) mode with no connection to any outside network. According to the company's documentation, in air-gapped mode all inference runs inside the customer's network, and not a single request goes out to Tabnine's own servers. This advantage is aimed mainly at sensitive organizations in finance, defense and government that cannot send proprietary code to an outside cloud.

A hard period: staff cuts

Like many other tech companies, Tabnine went through harder periods on the way to its acquisition. At one point, amid growing competition from tech giants that launched competing coding tools of their own (such as Microsoft's GitHub Copilot), the company laid off about 18% of its staff as part of an efficiency move. Periods like this are common in a competitive market where tools based on huge language models, backed by the enormous computing resources of giant companies, compete directly with smaller independent companies for exactly the same developer audience. The fact that Tabnine kept operating and was finally sold in a deal, rather than shut down or wiped off the map, shows that its distinct privacy positioning kept finding customers of its own, even against this competition.

2026: the acquisition by Tricentis

On 30 July 2026, Tricentis, a provider of automated software testing solutions for enterprises, which is not Israeli and not originally an AI company, announced that it was acquiring Tabnine. According to Calcalist, the deal was valued at tens of millions of dollars, though the parties did not publish an exact official price. Tricentis's official announcement said that Tabnine's Enterprise Context Engine technology would be built into its agentic testing platform. The goal is to give AI agents that test software a deeper understanding of the enterprise systems they test, before they decide what to test and how. According to the acquisition announcement itself, existing Tabnine customers will continue to receive support for current products as usual.