Matteo Collina, Node.js Technical Steering Committee member and co-founder of Platformatic, has proposed adding a first-class Virtual File System (VFS) to Node.js core via a new node:vfs module. The proposal, submitted as PR #61478 on the Node.js GitHub repository, introduces approximately 19,000 lines of code across 100 files and has ignited significant community discussion, both around the feature itself and the fact that it was largely built using Claude Code.
Node.js has always centred its identity around I/O, with streams, buffers, sockets and files forming the backbone of the runtime. However, common workflows such as running AI-generated code, executing tests without touching the disk, bundling Single Executable Applications (SEAs) and sandboxing file access in multi-tenant platforms have historically required workarounds. The proposed node:vfs module aims to address all four by providing an in-memory, fs-compatible API with mount points, overlay mode, symlinks and module loading hooks.
Alongside the core proposal, Platformatic has released @platformatic/vfs as a userland package on npm, available today for Node.js 22 and above. The package mirrors the API proposed for core, and when node:vfs ships, migrating is a one-line import change: swap '@platformatic/vfs' for 'node:vfs'. Vercel has independently extracted the same API into its own node-vfs-polyfill package, and LangChain has already adopted VFS for agent sandboxing through @langchain/node-vfs.
The feature itself has drawn support from developers who see clear benefits. On Hacker News, one commenter noted that:
Yarn, pnpm, webpack all have solutions for this. Great to see this becoming a standard. I have a project that is severely handicapped due to FS. Running 13k tests takes 40 minutes where a virtual file system that Node would just work with it would cut the run time to 3 minutes.
On Hashnode, developer Jake Morrison highlighted the AI agent use case:
I've been writing generated code to temp files and then cleaning up after and it always felt wrong. Keeping it in memory and importing directly makes so much more sense.
However, the PR has become a lightning rod for a broader debate. Collina was transparent that the implementation was built with significant AI assistance over Christmas 2025, stating that he "pointed the AI at the tedious parts" while focusing on architecture, API design and reviewing every line. In response, Fedor Indutny, a long-time Node.js core contributor known for initiating the io.js fork, launched a petition on GitHub calling for AI-generated code to be disallowed in Node.js core. The petition argues that the scale of the PR raises concerns around the Developer Certificate of Origin (DCO) and the ability of reviewers to adequately verify machine-generated code. On Reddit, Indutny wrote:
I firmly believe that the critical infrastructure Node.js is, is not the place for such changes.
Others pushed back, with some Hacker News commenters pointing out that "if there's anyone I would trust in exploring these avenues, it's him and the maintainers doing god's work in the Node.js repo in these past few years."
Neither Bun nor Deno currently offer an equivalent built-in VFS. Deno has already opened issue #32783 to track adding node:vfs compatibility. Bun, which leads benchmarks in raw file system throughput, has not announced comparable plans. The Node.js TSC is expected to vote on policies governing AI-assisted contributions, which will determine the fate of the PR. Documentation for the userland package and migration path is available on the Platformatic blog and the @platformatic/vfs GitHub repository.