jumpyBrain is not published to npm yet. The current public install path is a shell installer that clones/builds the app locally, initializes memory, verifies QMD, and installs detected agent integrations.
curl -fsSL https://raw.githubusercontent.com/nikoatwork/jumpyBrain/main/install.sh | bashDefault behavior:
- installs the app under
~/.jumpybrain/app - creates a CLI shim at
~/.jumpybrain/bin/jumpybrain - initializes machine-global Markdown memory at
~/.jumpybrain/memory - verifies or installs QMD for local indexing/recall
- installs all detected integrations:
- Codex:
~/.agents/skills/jumpybrain-memory/SKILL.md - Claude Code:
~/.claude/skills/jumpybrain-memory/SKILL.md - Pi:
~/.pi/agent/extensions/jumpybrain-memory.ts
- Codex:
Add the bin directory to your shell if you want jumpybrain everywhere:
export PATH="$HOME/.jumpybrain/bin:$PATH"Then verify:
~/.jumpybrain/bin/jumpybrain doctor
~/.jumpybrain/bin/jumpybrain recall --root ~/.jumpybrain/memory --topic "what should I remember?" --limit 5Use project scope when you want memory and skills in the current repository instead of global locations:
curl -fsSL https://raw.githubusercontent.com/nikoatwork/jumpyBrain/main/install.sh | bash -s -- --scope project --integrations allProject scope creates ./memory and installs project-local integrations:
- Codex/Pi portable skill:
.agents/skills/jumpybrain-memory/SKILL.md - Claude Code skill:
.claude/skills/jumpybrain-memory/SKILL.md - Pi extension:
.pi/extensions/jumpybrain-memory.ts
--scope global|project Default: global
--memory-root <path> Override the memory root
--integrations auto|all|none Default: auto; auto installs detected harnesses
--ref <git-ref> Install from a branch/tag/commit ref
--source <path-or-git-url> Install from a local checkout or alternate git URL
--install-root <path> Default: ~/.jumpybrain
--dry-run Print planned actions
Use --integrations all to install every integration regardless of detection. Use --integrations none for CLI-only setup.
Uninstall removes installer-owned app, shim, manifest, and integration files. It preserves Markdown memory by default.
curl -fsSL https://raw.githubusercontent.com/nikoatwork/jumpyBrain/main/uninstall.sh | bashTo intentionally remove the configured jumpyBrain memory root as well:
curl -fsSL https://raw.githubusercontent.com/nikoatwork/jumpyBrain/main/uninstall.sh | bash -s -- --delete-memory--delete-memory refuses broad/unowned paths and requires a jumpybrain.json memory config.
jumpyBrain is intentionally QMD-first. The installer verifies qmd --version; if QMD is missing it tries:
npm install -g @tobilu/qmdIf your npm global install location is not writable or not on PATH, install QMD manually or set:
export JUMPYBRAIN_QMD_BIN=/path/to/qmdResolution order for local/server runtime use:
JUMPYBRAIN_QMD_BIN- bundled/package-local
node_modules/.bin/qmdwhen present qmdonPATH
QMD-derived files live under <memory-root>/.jumpybrain/ and can be rebuilt from canonical Markdown.
Remember writes memory; recall reads memory:
~/.jumpybrain/bin/jumpybrain recall --root ~/.jumpybrain/memory --topic "current task" --limit 5
printf '%s\n' "Markdown remains canonical; indexes are rebuildable." \
| ~/.jumpybrain/bin/jumpybrain remember --root ~/.jumpybrain/memory --type decision --title "Memory storage rule"
~/.jumpybrain/bin/jumpybrain recall --root ~/.jumpybrain/memory --query "Where is the memory storage rule?" --limit 5 --jsonWhen running inside a repo initialized with memory/jumpybrain.json, agents can use recipe shortcuts that discover the root:
jumpybrain run memory:recall --topic "memory storage" --limit 5If you manually add or edit Markdown memory files, run:
jumpybrain index --root <memory-root>For development from a checkout:
npm install
npm run build
npm link
jumpybrain --helpFor dogfooding in another repo, maintainer local tarball workflows are documented in local-cli-builds.md.
jumpyBrain is structured around three paths while keeping one memory system:
- Local runtime install: current installer path; CLI and runtime run on the same machine as a local Markdown memory root.
- Hosted client install: future thin CLI client pointed at a deployed jumpyBrain server. Remote targets are recognized in the CLI but not implemented yet.
- Server deploy: future/self-hosted runtime on a VPS/server against a server-local Markdown memory root.
Markdown remains canonical in all paths; indexes and support state remain derived/rebuildable.