Monolithos: an AI workspace built around a memory vault you hold

Monolithos is an AI workspace organised around persistent personal materials and memory. Its official site describes a vault based on Markdown and YAML, shared by tools for recall, multi-step tasks, writing, presentations and audio. The central proposition is continuity: work should be able to use context you have already built.

Keep the material readable and editable

The publisher says users hold their memory and vault, and can inspect or edit the underlying files. It also describes native access to an existing Obsidian vault without an import or duplicate database. That is relevant if your knowledge already lives in notes you want to keep independently readable.

Memory changes are not presented as automatic authority over your profile. The site says its Watcher proposes changes for confirmation, and recall includes sources. A useful trial would inspect one proposed memory, correct it and check how a later answer uses the revision.

Distinguish data ownership from local processing

Monolithos explicitly says ownership does not mean all computation happens locally. Cloud AI processes the context needed for a request, while the publisher describes private-domain material as excluded from cloud AI context. Those are product descriptions, not an independent privacy audit.

Before connecting sensitive material, review the actual settings and data policy for the model path you intend to use. Holding readable files and choosing what reaches a model are related but separate parts of the workflow.

Separate the current app from the roadmap

The site lists current memory and AI work capabilities, but says the complete device-off continuing-agent experience is still in development. Its robot and medical digital-twin sections also carry testing or development labels. Do not treat those demonstrations as capabilities available in the ordinary app download.

For a first evaluation, use a small vault and one task that draws on it. Check the answer's sources, any proposed changes and the files left behind. Monolithos makes the most concrete sense as a workspace for maintaining context today; its broader personal-agent ambitions should be judged as they reach actual releases.