First useful path

Getting started

Try AIppocampus locally before you configure anything ambitious: probe the CLI, check provider status, onboard only with consent, then search one old phrase from source.

The short version

  1. Use Python 3.12 or newer.
  2. Probe the packaged CLI without writing memory artifacts.
  3. Check local conversation providers in read-only mode.
  4. Register local history only after explicit consent.
  5. Search for one distinctive old phrase and reopen source when claims matter.
Install guide README first checks

Commands

Start with read-only checks:

uvx aippocampus --help
uvx aippocampus onboard --provider codex --status --format json

After you decide this machine may create local memory artifacts:

uvx aippocampus onboard --provider codex --all --format json
uvx aippocampus search "a distinctive old phrase"

What to expect

It feels like preparing a library

AIppocampus scans local source, builds clean source, and prepares searchable indexes. It is not a hosted chat memory switch.

The first payoff is small

A useful recall result points back to earlier source. The agent can continue from evidence instead of acting as if it magically remembers.

Hooks come later

Manual search proves the substrate. Prompt and lifecycle hooks make future continuity quieter after you trust the local setup.

Privacy stays first

Pause before onboarding if the workspace contains material you do not want in local memory artifacts.

Next stops

Read the architecture page for the mental model, then the agent page if you are wiring AIppocampus into an agent host.

Architecture For agents

Adapted from the public wiki pages Getting Started and First Recall and Onboarding.