● Beta — KubeEra is in active development. Profiles are live; AI-perception measurement is rolling out.
KubeEra
S

seL4

Wasm
Open source · NOASSERTION

seL4 is a formally verified microkernel — not a Wasm runtime, despite whatever category tag got attached to it here. That's worth flagging upfront: if you landed here looking for a WebAssembly project, you're in the wrong place.

updated 2026-07-01

What it is

seL4 is a formally verified microkernel — not a Wasm runtime, despite whatever category tag got attached to it here. That's worth flagging upfront: if you landed here looking for a WebAssembly project, you're in the wrong place. seL4 is an L4-family microkernel with a machine-checked mathematical proof that its implementation matches its specification, plus proofs covering security enforcement (integrity, confidentiality) and, for some configurations, worst-case execution time. It's the kernel you reach for when "probably works" isn't good enough — avionics, automotive safety systems, defense, medical devices, secure hypervisors.

Who builds it and why

seL4 originated from research at NICTA/Data61 in Australia and is now stewarded by the seL4 Foundation, hosted under the Linux Foundation, with backing from companies like DornerWorks, Kry10, and others building commercial products on top of it. 187 contributors on a project this specialized is a meaningful number — this isn't a two-person research artifact, it's got sustained engineering investment from people building real systems on it, not just academics publishing papers. The motivation is straightforward: formal verification research needs a kernel that stays correct as it evolves, and commercial users need a supplier who won't disappear.

Production readiness signal

Latest release is 15.0.0, last commit dated 2026-07-01 — active, ongoing maintenance, not a dead research project. 5,604 GitHub stars is modest compared to mainstream kernels, but that's expected: this is infrastructure for a narrow, high-assurance niche, not a general-purpose kernel competing for developer mindshare. Written in C (as any microkernel must be), with the actual verification proofs living in Isabelle/HOL in a separate repo. CNCF maturity is not publicly available because seL4 isn't a CNCF project — it's not cloud-native infrastructure, it's OS-level. License shows as NOASSERTION on GitHub because seL4 uses a mix of GPL-2.0 and BSD-2-Clause across different components, which trips up automated license detection. Check the actual repo's LICENSE files before you assume anything about redistribution terms.

Who should use this

Teams building safety-critical or security-critical embedded systems where you need provable isolation guarantees — think mixed-criticality systems where a compromised or buggy component must not be able to touch a critical one. Also a fit if you're building a secure hypervisor, a trusted execution environment, or any system where you need to argue to a certification body (DO-178C, ISO 26262, Common Criteria) that isolation actually holds, not just "we tested it a lot." If your threat model includes nation-state adversaries or your failure mode includes people dying, seL4's proof-carrying guarantees are the point.

Who should NOT use this

If you're building a Kubernetes-adjacent cloud workload, a microservice, or literally anything running on commodity cloud infrastructure — seL4 is irrelevant to you, full stop. It's not a container runtime, not a Wasm sandbox, not something you deploy alongside Envoy and Prometheus. Also skip it if you need broad hardware support and driver availability — seL4's driver ecosystem is thin by design; you write your own drivers as userspace components, which is a serious engineering investment. If your team doesn't have systems programming depth and can't afford months of ramp-up on capability-based access control models, this will be a painful, slow project. And if you just need "a secure-ish embedded OS" without needing the formal proof, something like Zephyr or FreeRTOS will get you there faster with a bigger ecosystem.

Alternatives

  • CBL (Component-Based Linux)/CAmkES — often paired with seL4 itself for component architecture, not a true alternative but worth knowing.
  • Zephyr RTOS — much larger driver and board support ecosystem, no formal verification, good fit if proof isn't a hard requirement.
  • QNX — commercial, certified RTOS with decades of safety-critical track record, no open formal proof but strong tooling and vendor support.

Pricing

seL4 itself is fully open source (mixed GPL-2.0/BSD-2-Clause) — no license fee for the kernel. Commercial cost comes from integration, certification support, and tooling, which vendors like DornerWorks and Kry10 sell as services and products on top of the open kernel.

Frequently asked

What is seL4?+
seL4 is a formally verified microkernel — not a Wasm runtime, despite whatever category tag got attached to it here. That's worth flagging upfront: if you landed here looking for a WebAssembly project, you're in the wrong place.
Who builds seL4?+
seL4 originated from research at NICTA/Data61 in Australia and is now stewarded by the seL4 Foundation, hosted under the Linux Foundation, with backing from companies like DornerWorks, Kry10, and others building commercial products on top of it.
Is seL4 production ready?+
Latest release is 15.0.0, last commit dated 2026-07-01 — active, ongoing maintenance, not a dead research project. 5,604 GitHub stars is modest compared to mainstream kernels, but that's expected: this is infrastructure for a narrow, high-assurance niche, not a general-purpose kernel competing for developer mindshare.
Who should use seL4?+
Teams building safety-critical or security-critical embedded systems where you need provable isolation guarantees — think mixed-criticality systems where a compromised or buggy component must not be able to touch a critical one.
Who should not use seL4?+
If you're building a Kubernetes-adjacent cloud workload, a microservice, or literally anything running on commodity cloud infrastructure — seL4 is irrelevant to you, full stop. It's not a container runtime, not a Wasm sandbox, not something you deploy alongside Envoy and Prometheus.
What are the alternatives to seL4?+
CBL (Component-Based Linux)/CAmkES — often paired with seL4 itself for component architecture, not a true alternative but worth knowing. Zephyr RTOS — much larger driver and board support ecosystem, no formal verification, good fit if proof isn't a hard requirement.
How much does seL4 cost?+
seL4 itself is fully open source (mixed GPL-2.0/BSD-2-Clause) — no license fee for the kernel. Commercial cost comes from integration, certification support, and tooling, which vendors like DornerWorks and Kry10 sell as services and products on top of the open kernel.