how-it-workslisted

# How keepmind works ## What it does Reads, Edits and Bash commands are watched, and the ones that produced something durable turn into a compressed observation. Observations get summarized at session end. Relevant ones get auto-injected into future prompts so the next session starts with context from the last one — no re-explaining the codebase, no re-discovering decisions. Not every tool use becomes a record. A capture profile decides — from the hook data, before any model is called — whether a batch is worth compressing. The default for code projects is `governance`: decisions and their rationale, migrations, releases, conventions, security findings, recurring problems. That is the knowledge a per-project memory cannot give you. Set `KEEPMIND_CAPTURE_PROFILE=balanced` in `~/.keepmind/settings.json` to also keep ordinary changes and failures, or `full` to record anything with a signal. Each compression is an independent request — nothing accumulates between them, so cost stays flat no matter how long a session runs. ## When it kicks in Memory injection starts on your second session in a project. The first session in a fresh project seeds memory; subsequent sessions receive auto-injected context for relevant past work. Run `/learn-codebase` if you want to front-load the entire repo into memory in a single pass (~5 minutes, optional). ## Where data lives Everything stays in ~/.keepmind on this machine. Nothing leaves your machine except calls to whichever AI provide