At PayFit, the engineers had already voted with their feet. Their company had put teams into a shared software repository without enough support. Teams left. Independence was worth the inconvenience of shipping a feature through several repositories, even when deployment took two to five days. A supposedly efficient arrangement had become something people wanted to escape.
- Nx maps connected projects, then skips work a change cannot affect.
- The open-source tools are free; Nx Cloud sells shared caching, distributed CI and enterprise controls.
- Adoption needs ownership and support. A migration order cannot supply either.
PayFit’s four-person platform team changed the proposition. It repaired the workspace, configured Nx task inference, adopted Nx Cloud and brought back pilot teams. Colleagues could see a functioning alternative. Nx’s account says deployment eventually reached a maximum of 20 minutes. The lesson is wonderfully unglamorous: make the shared thing useful, and people may choose to share.
01 A map before a motor
Nx makes tools for an awkward stage of software growth: the moment when changing one small thing means checking a bewildering number of other things. A checkout application borrows a button library. That library borrows design tokens. Somewhere, a developer changes a color, and a build pipeline starts behaving as though the entire company needs an inspection.
A monorepo puts multiple projects in one repository. Nx adds knowledge about their relationships. Its project graph records dependencies; its task graph organizes the jobs needed to build and test them. Affected analysis narrows the work. Caching restores previous results when the relevant inputs match. Nx Cloud lets colleagues and CI machines share those results.
dependencies
changed
already done
The expertise here is build orchestration: understanding dependencies, scheduling work and preserving useful results. Nx is not the compiler that turns every language into executable code. It coordinates the tools teams already use, with plugins that add deeper knowledge of frameworks and languages.
02 Google habits, smaller-company packaging
Co-founders Jeff Cross and Victor Savkin worked at Google and on Angular. They had seen large-scale monorepos with the infrastructure required to make them practical. Their company, started in 2016, brought that thinking into open-source tooling. Cross is CEO; Savkin is CTO. In September 2023, Nx announced a $16 million Series A with Nexus Venture Partners and a16z.


That background explains the product’s breadth. Generators create code that follows conventions. Migrations help update workspaces. Module boundaries restrict which projects may depend on others. Nx Console brings workspace exploration into the editor. Nx Labs supplies training and engineering help when the missing ingredient is experience rather than another command.
The company’s working arrangements fit its developer audience. Its careers page describes flexible locations and hours, with colleagues spread across North America and Europe. It also describes an unusual internal policy: employees know what everyone earns. Transparency is a declared operating practice, down to the compensation spreadsheet.
03 Forty-five minutes is a long coffee
Customers include platform teams maintaining software for payroll, cloud hosting, accounting and broadcasting. These are organizations where shared infrastructure can become everybody’s problem and nobody’s favorite assignment.
Consider Hetzner Cloud. Its GitLab merge trains repeatedly ran builds and tests on unchanged code. The team evaluated Nx and Turborepo, chose Nx, and combined modularization with remote caching and distributed execution. Its published case study reports typical pipelines falling from 45 minutes to around six, although some took up to 20.
Caseware offers another scale test: more than 700 projects and a reported 93% remote-cache hit rate. Its migration expanded from shared frontend libraries into backend applications, supported by a dedicated migration team. The human infrastructure mattered alongside the software.
04 The price of remembering
Nx’s business model follows that split. The open-source build system is free. Nx Cloud charges for the shared infrastructure and services around it. On October 15, 2026, its Hobby tier included 50,000 monthly credits and five contributors. Team started at $29 a month, with additional contributor, usage and concurrency charges. Enterprise pricing was custom.
The comparison belongs beside the engineering bill. Turborepo, Bazel and Rush offer alternative approaches to build coordination. Nx’s proposition combines a graph, framework-aware plugins, code generation, architectural rules and a commercial CI layer. Lerna is a special case: Nx took over its stewardship in 2022, and modern Lerna uses Nx for task execution.
For a small application with cheap tests, another orchestration layer may save little. Caching also depends on correctly describing inputs; a hidden environmental dependency can undermine reuse. Buyers should measure representative workflows, include migration effort and inspect the total cloud bill. A fast demonstration is only the beginning of the calculation.
05 More autonomy needs better boundaries
The trust question became painfully concrete in August 2025. Attackers exploited a GitHub Actions injection vulnerability, stole an npm token and published malicious Nx packages. Nx’s postmortem says the packages were available for four hours and that Nx Cloud was unaffected. The company moved to npm Trusted Publishing and introduced stricter release approvals. A build tool sits close to valuable credentials.
Meanwhile, Self-Healing CI, introduced in June 2025, examines failed tasks, proposes repairs and validates them. The graph supplies codebase context. Developers can review fixes; configured rules can allow automatic application. The ambition is to shorten the whole repair cycle, including the human interruption after a red build.
“Making developers more productive”Nx’s stated focus
That direction continues in October 2026’s Nx 23.3: more precise, opt-in task selection and output designed for coding agents. The practical idea readers can borrow is older than AI: map the dependencies, remove repeated work, prove the improvement with a pilot. Then make it easy for the next team to join.