Skip to content

Releases

v0.2.1 - claude.ai support via brain CLI

Choose a tag to compare

@jim4226 jim4226 released this 23 May 19:33

Adds support for claude.ai users (no Claude Code required).

Added

  • brain.py unified CLI with subcommands show, copy, lint, serve, eval.
    • brain copy writes your brain to the clipboard (Windows clip / macOS pbcopy / Linux xclip/wl-copy/xsel).
    • Paste into any AI chat at session start.
  • README section: "Use with claude.ai (no Claude Code required)" documenting two patterns:
    • Pattern A — upload WORKSPACE_BRAIN.md as claude.ai Project knowledge so every chat in that project sees it.
    • Pattern B — brain copy + manual paste for ad-hoc sessions.
  • Honest about what's lost without hooks (no automatic flush) vs what's kept (file structure, linter, viewer, eval).

Why

The hooks are Claude-Code-specific. The brain file format, the linter, the HTML viewer, and the eval harness aren't. This release lowers the activation energy for claude.ai users to get most of the value without changing tools.

Usage

git clone https://github.com/jim4226/claude-workspace-brain
cd my-project   # where your WORKSPACE_BRAIN.md lives
python /path/to/claude-workspace-brain/brain.py copy
# now paste into claude.ai

Or use claude.ai Projects: upload WORKSPACE_BRAIN.md as Project knowledge once, re-upload when it changes.

v0.2.0 - Viewer + Eval + /brain-init + /brain-archive

Choose a tag to compare

@jim4226 jim4226 released this 23 May 18:12

The "extraordinary first impression" release.

Added

  • template/brain-viewer.html — self-contained interactive viewer with dark/light themes, animated force-directed synapse graph, section cards. Renders any WORKSPACE_BRAIN.md served over python -m http.server.
  • eval/abtest.py — A/B harness using the Anthropic SDK. Asks Claude 5 brain-dependent questions with/without the brain, judge model scores both blind 0-30, reports mean delta + 95% CI. Prompt caching cuts brain cost ~90%. --dry-run validates without API calls.
  • /brain-init slash command — 6-question Q&A flow that scaffolds a personalized brain in 60 sec.
  • /brain-archive slash command — auto-migrates stale completed threads + old sessions to WORKSPACE_BRAIN_ARCHIVE.md, with approval gate.
  • CHANGELOG, CONTRIBUTING, SECURITY, issue + PR templates.
  • CI workflow: smoke tests on Linux / macOS / Windows x Python 3.8 / 3.11.

Changed

  • README: new hero with two-line tagline, embedded viewer screenshot, mermaid lifecycle diagram, sharpened comparison table vs Cipher / mem0 / memory-compiler / MemU.
  • Brain viewer accepts ?theme=dark|light URL param.

Fixed

  • brain_lint.py no longer crashes on emoji in section titles (UTF-8 stdout reconfigure on Windows).

v0.1.0 - Initial public release

Choose a tag to compare

@jim4226 jim4226 released this 23 May 18:12

The minimum viable claude-workspace-brain.

What's in this release:

  • SessionStart hook injects WORKSPACE_BRAIN.md at session start
  • PreCompact hook directs Claude to flush state before context compaction
  • Optional Stop hook for opportunistic flushes (off by default)
  • /brain and /brain-grade slash commands
  • brain_lint.py — 6-axis static quality scorer (100 pts), zero deps
  • Cross-platform installer (install.sh, install.ps1) with idempotent settings.json merging
  • Generic template + research / web-app examples

Install:

curl -sSL https://raw.githubusercontent.com/jim4226/claude-workspace-brain/main/install.sh | bash

(macOS / Linux. PowerShell users: see README for the irm one-liner.)