VectraOS is built by people who think the datacenter has no business owning your
computer. The reference port is pure Rust, and that is where the work is. If you can write
#![no_std] and read a page-table entry without flinching, we need you.
Contribution channels are still being stood up. The repository is invited-collaborator only until it goes public; formal governance is a draft, not in effect until V1 ships. What's below is how it will work — and the posture that won't change.
Pre-V1, VectraOS is a BDFL project — small, opinionated, and conservative by design. The point of a benevolent dictator is to transition out of one as the project grows, not to avoid ever being a real community. The rules were written early, while they're easy to write.
The kernel is the deep end and the most urgent need. But the climb from substrate to a livable machine is long, and almost all of it is Rust.
no_std microkernel work on aarch64 and x86_64 — scheduling, SMP, paging, the syscall surface, capability enforcement. Zero third-party crates; you read everything.
Drivers are userland Tasks holding DeviceCap / IrqCap / DmaCap. NIC, storage, display — across three deliberate strategies, detailed below.
One Task, one Ring schema: vfsd, netd, tlsd, compositord, and the rest. A signal-mask event loop and a declared capability footprint.
The small no_std crates everything else stands on — text, time, crypto, encoding, the cooperative executor. Get these right and the whole stack gets easier.
Data-oriented GUI, graphics, storage, documents — then the editor, terminal, mail client. Further out, but real Rust work for people who'd rather live above the kernel.
Reproducible-build tooling, QEMU and on-hardware testing, documentation, and a forthcoming code of conduct. Drive-by contributions are welcome and valued.
Not a Rust developer? There's still room — the spec, the distribution and trust design, hardware bring-up, and the docs all need hands. But be honest with yourself about the center of gravity: this is a Rust operating system, and most of the open work is systems programming.
A driver is a Task that wants a DeviceCap. How we get there depends on what already exists and under whose license. The base image you boot ships zero GPL binaries.
Where the protocol is small and open — buses, UART, timers, virtio, interrupt controllers — we write the driver fresh in Rust, straight from the public datasheet. The source of truth is the spec, not somebody else's code.
Where a mature driver already exists under a permissive license, we compile it unmodified against a thin Rust KPI shim. The four BSDs — FreeBSD & DragonFly (BSD-2), NetBSD (BSD-2), OpenBSD (ISC) — are a bigger, link-compatible, GPL-free pool than any one alone. OpenBSD's clean-room Wi-Fi tree drags modern wireless out of the GPL air-gap entirely.
Where only a GPL driver reaches the hardware — in practice, modern GPU modesetting — it runs behind the Linux-driver personality, shipped as a separate, opt-in package paired with its corresponding source. Your license stays clean unless you choose to cross that line yourself.
The canonical, per-driver bring-up plan — every tier, every
chipset, and the license boundary enforced by a CI gate at code-review time — lives in the
Swift reference port's design docs (docs/forward/kernel/driver-strategy.md).
The Rust port follows the same three-tier policy, with Rust standing in for Swift on the native tier.
Argue about how to build it — that's the work. But the foundational vision is not up for committee revision. Contributions are expected to align with the refusals that make this project worth doing:
Disagreement on the foundational vision means you want a different project — and that's fine. Forks are a feature.
Every file carries an SPDX header. You retain copyright on what you write. No contributor license agreement, no copyright assignment. The OS is free in full — no "Pro Edition," no paywalled features. BSD-3 over BSD-2 for one reason: nobody gets to use the VectraOS name to endorse their own derivative.
Vision-aligned funding is welcome — paying core devs to work full-time is a goal, not a threat. But sponsors don't buy board seats, roadmap items, or technical direction. The source of money stays separate from the source of decisions.
BSD-3 permits them; the project welcomes them. A fork that disagrees with the BDFL is legitimate — it just can't use the VectraOS name to endorse itself, per the licence's third clause. Take the code and go build your version.
AI assistance is allowed with human review. The human is the contributor of record and accountable for the code; commits note the agent in a footer. AI authorship doesn't accrue toward core-contributor status — humans only.
Public comment on governance is closed until first beta — the project is still too small for committee-by-comments to produce useful signal rather than dilute an opinionated vision. The full model, including the "catches fire" clause that transitions the project to community governance once it's big enough, lives in the governance RFC.
A project that hands the user back their computer cannot itself be captured by a megacorp.
Channels for expressing interest go here once the repository is public. Until then: read the architecture, study what's already running, and come ready to talk before you code.