AetherMind is an open source continuity primitive for AI-assisted work.
It gives agents a small, shared way to preserve load-bearing context beside the project they are changing: decisions, corrections, uncertainty, friction, and verification that should survive the current chat or agent run.
It is not a transcript store, task log, project manager, or hosted memory service.
AI coding agents lose useful working judgment across context limits, handoffs, fresh sessions, and tool boundaries. AetherMind records that judgment as append-only layers in a project-local .aethermind/ store.
Any agent that can run shell commands can use it.
aethermind init --root . --purpose "start project continuity"
aethermind layer --root . --type decision --ctx build --body "Use the CLI as the universal v0.1.0 surface"
aethermind validate --root .init creates .aethermind/ and writes an init layer. layer records durable work context. validate, status, and inspect are diagnostic commands; they are not a substitute for writing continuity during real work.
For work on systems where you should not leave project artifacts by default, write local remote-work continuity:
aethermind remote-note \
--remote customer-host:/srv/app \
--task incident-1 \
--ctx triage \
--body "Inspected service state; kept continuity locally"The remote target is metadata. The note is stored locally under a remote-work store and does not create .aethermind/ on the remote/customer system by default.
Basic recipes for Codex, Claude Code, and Grok Build live under contrib/recipes/. They are examples of using the CLI from agent runners; they are not separate architectures.
See SCOPE.md for the product boundary and what belongs in an AetherMind layer.