What it is
Memory Utility Labs is the imprint over a small body of local-first systems work: agent infrastructure, safety substrates, and evidence tooling. Each project stands on its own repository with its own documentation; the Labs name means it was built to the same doctrine — provenance on every claim, receipts for every consequential act, and honest limits stated in the open.
Published under it
- aegis — agent operating system for AI coding work. v1.2.0, public on GitHub. See the AEGIS volume.
- hermes-refuse — controlled execution for macOS agents. Architecture and design stage, Apache-2.0. Not production-ready, and not presented as such.
- interlock — agent safety substrate. v0.1.0, MIT, CI green.
- reclamation-evidence-ledger — provenance-stamped environmental evidence from Sentinel-2. MIT.
Publications
Introducing AEGIS — Technical Specifications, Vol. I (2026). The publication of record for the AEGIS system: doctrine, data plane, and the honest-yield rule, set down before the implementation had to answer for it.
Honest status
The Labs' charter is still being written. What exists today: the imprint, the publication series, and the projects published under it. Anything beyond that would be a claim without a receipt, so it is not made here.