Teaches AI agents how to use HEBBS, a local-first cognitive memory engine that stores, indexes, and retrieves knowledge across four dimensions.
git clone https://github.com/hebbs-ai/hebbs-skill.git ~/.openclaw/skills/hebbsRestart your OpenClaw session to pick up the skill.
git clone https://github.com/hebbs-ai/hebbs-skill.git ~/.claude/skills/hebbs- Download this repo as a ZIP
- Go to Settings > Capabilities > Skills > Upload the ZIP
HEBBS creates a .hebbs/ directory next to your files: a self-contained cognition layer. Build the index once, then share it across agents, machines, or your team. Everyone gets the same memory instantly.
.hebbsignore works like .gitignore: your private files stay private, your agents only see what you allow.
Your files are the source of truth. .hebbs/ is derived and rebuildable. Delete it anytime and run hebbs init . && hebbs index . to get it back.
macOS (Homebrew):
brew install hebbs-ai/tap/hebbsAny platform (Linux, macOS):
curl -sSf https://hebbs.ai/install | shhebbs init . --provider openai --key $OPENAI_API_KEY
hebbs index .
hebbs recall "your query here"--model is optional (defaults per provider). Embedding auto-configures when using OpenAI. The daemon starts automatically and watches for file changes.
hebbs statusThe tune/ directory contains a skill for agent-driven retrieval tuning: generate evals, run baselines, optimize parameters, and export compiled rules.