Coding assistants forget everything from past sessions. Your AI conversations produce a steady stream of project-specific knowledge (conventions, gotchas, named modules, decision rationale), and most of it evaporates when the session ends. kenkeep captures it, asks a human to curate it, commits it to the repo, and injects it back into every future session — on Claude Code, Codex CLI, Cursor, OpenCode, and GitHub Copilot CLI.

Why kenkeep

Built up and shared across your team

The knowledge base grows in your repo as plain markdown, one node per fact, accumulated from real coding sessions. It travels with the project through git pull, so every teammate works from the same conventions instead of rediscovering them on their own laptop.

Reviewed and versioned like code

Nothing reaches the knowledge base without a human approving it. Every addition or change is an ordinary git diff you review in a commit or PR, with the full history there to inspect, blame, or revert like any other code.

No extra infrastructure

No daemons, services, databases, or vector stores. kenkeep is just Node and git, so there is nothing to provision, host, or keep alive, and nothing new to secure.

No API keys

It all runs from within the assistant of your choice, on the subscription you already pay for. There is no separate API key to obtain, store, or rotate.

How it works

kenkeep runs a loop around your AI sessions. Capture and injection happen on their own; you trigger curation, and you decide what to keep:

• Capture (automatic): when a session ends, a hook saves the transcript.
• Curate (you run /kk-curate): the AI drafts proposed notes under nodes/, then walks you through any contradictions with an existing note.
• Review (you decide): inspect the notes with git diff, then commit the ones you want to keep. INDEX.md is injected into every future session.

Full walkthrough: How it works.

Quick start

npx kenkeep init --harnesses claude
npx kenkeep doctor

Swap claude for codex, cursor, opencode, or copilot (or pass a comma-separated list). For GitHub Copilot CLI, npx kenkeep init --harnesses copilot installs the skills under .github/skills/ (Copilot's documented project skill location) and keeps the adapter's hook scripts under the project-local .copilot/ directory, registering them in the user-level ~/.copilot/hooks/kk.json.

Then code normally. When you want to turn captured material into knowledge nodes, run /kk-curate inside your harness session (also /kk-add, /kk-bootstrap). The skills are context-aware and walk you through conflict resolution. New nodes appear in nodes/; review with git diff and commit the ones you want to keep.

Seed from existing docs

If your repo already has READMEs, ADRs, or module docs, seed the knowledge base from them. Inside a harness session:

/kk-bootstrap

The scan walks the repo root, filtered by .kkignore (generated by init, uses gitignore-style syntax). Edit .kkignore to exclude directories you don't want scanned. Review the resulting nodes under nodes/ with git diff and commit the ones you want to keep.

Add knowledge manually

At any time during a session you can use /kk-add to make sure the assistant remembers a message. Just casually mention it, and you're done:

No, you got that wrong.

This project aims to maximize code re-use, instead of duplication. Adapt and extend
the abstractions to fit this use case. Also, /kk-add this.

Documentation

Full documentation: https://mateuaguilo.com/kenkeep

For maintainers of this package itself, see CONTRIBUTING.md.

License

MIT