nocktechnologies

Nockbrain — Development skill for Claude Code

Development community

Memory you approve for Claude Code: a small, signed, local store of reviewed facts, recalled when you ask about the past.

How to install Nockbrain

This entry records only its repository, not the path inside it, so there is no exact command to give. Open nocktechnologies/nockbrain and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

What Nockbrain does

Memory you approve for Claude Code: a small, signed, local store of reviewed facts, recalled when you ask about the past.

Alternatives in Development

Evolving Agents — A director agent that learns how your team works: it plans by walking your CLAUDE.md and memory as linked stat 453 ★
Funes — Durable, searchable memory of your past agent sessions 349 ★
Memory Question — Ask Claude's memory system questions about past work, decisions, and context 332 ★

README

NockBrain

I built this because I was tired of explaining my own decisions to Claude Code again every few sessions, and I didn't trust a memory that writes itself. NockBrain keeps a small, signed store of facts you've reviewed, on your own machine. When a prompt asks about the past, it hands the matching facts back to Claude as reference. Nothing goes in until you've read it and applied it.

[![Watch the one-minute explainer](media/poster.png)](media/nockbrain-explainer.mp4)

The one-minute explainer, its captions and the script behind it are in [media/](media/).

Who it's for

NockBrain fits one person who works with Claude Code across many sessions and is tired of explaining the same decisions, rules and corrections again. It runs on macOS and Linux, and the stock macOS `python3` is enough. Claude Code gets automatic capture and recall; Codex CLI, Gemini CLI, Cursor and Claude Desktop can recall through the MCP server. If you want a memory that fills itself with no review step, this is the wrong tool.

How it works

You decide what goes in. Point NockBrain at notes you choose, or turn on an optional Claude Code hook that queues what was said in a session each time Claude finishes a reply. Either way the candidates land in a review queue with secrets redacted, and nothing is in memory yet. You read every candidate with its kind and the line it came from, then apply the batch or discard it. Applied facts are signed, and their signed content is never rewritten.

Recall is opt-in too. When a prompt asks ab