Open source · MIT licensed · No account required
Install ALIVE
Three minutes. Full context management.
Before you start
Prerequisites
Python 3.10+
python3 --version to check
A terminal
macOS, Linux, or Windows (WSL)
No accounts. No API keys. No databases. That's the whole list.
Install flow
Five steps, top to bottom.
Open Claude Code
Launch Claude Code in your terminal. If you don't have it yet, install it from Anthropic first.
Don't have Claude Code? Install it here
Install the plugin
Inside Claude Code, type /plugins and search for alivecontext/alive. Or run the install command directly from your terminal:
Exit and restart
Exit Claude Code, then open a new session. The plugin loads fresh on new sessions.
Set up your world
In the new session, run /alive:world. The squirrel walks you through creating your .alive/ folder, your first walnut, and your identity file.
Start working
That's it. Every session now loads your context automatically. Save with /alive:save when you're done. Your AI remembers next time.
Migrating from v1 or v2?
If you already have ALIVE installed, run /alive:system-upgrade instead of /alive:world. The upgrade migrates your existing context to v3 architecture. Nothing is deleted.
What just happened
Here's what's on your machine now.
.alive/ folder
Created in your world directory. This is where all context lives.
_kernel/ with three source files
key.md (identity), log.md (session history), insights.md (domain knowledge).
Computed files
now.json and tasks.json update automatically after each save. Scripts compute these from your source files.
Squirrel runtime active
15 skills, 13 hooks, 6 rules. Loaded in every Claude Code session automatically.
Python projections
project.py and generate-index.py run after each save to compute fresh state. That's why Python 3 is required.
First things to try
You're installed. Now explore.
/alive:worldSee everything at a glance/alive:load-context [walnut]Load a specific project/alive:saveCheckpoint your work/alive:capture-contextCapture external content/alive:search-worldFind anything across all your contextOr just chat naturally. The squirrel watches for decisions, tasks, and people mentions automatically. Everything routes on save.
FAQ
Common questions.
Do I need to pay for anything?
No. ALIVE is free and open source (MIT licensed). You need a Claude Code subscription ($20/mo from Anthropic) to use Claude Code itself, but the plugin is free.
Where does my data go?
Nowhere. Everything stays in plain markdown and JSON files on your machine. No cloud. No sync service. No telemetry. You can read every file in a text editor.
Does it work with models other than Claude?
The plugin runs in Claude Code, but the context files are agent-agnostic. The same .alive/ folder works with Hermes Agent, OpenClaw, and any tool that can read markdown. Ecosystem plugins for Obsidian, Claude Desktop, and Codex are in testing.
What if I already use CLAUDE.md files?
ALIVE works alongside CLAUDE.md. Your existing CLAUDE.md stays. ALIVE adds the lifecycle layer -- session saves, projections, task routing, cross-project context. Think of CLAUDE.md as static config, ALIVE as the living layer on top.
Why does it need Python?
Python runs the projection scripts (project.py, generate-index.py) that compute now.json after each save. These aggregate all your source files into a single snapshot. Python 3.10+ required. The rest of the system is plain files + bash hooks.
Can I use it for non-coding work?
Yes. ALIVE manages context for anything -- client projects, personal goals, research, people relationships, experiments. The ALIVE framework (Archive, Life, Inbox, Ventures, Experiments) covers your whole life, not just code.
What's a walnut?
A walnut is the unit of context. Any meaningful thing with its own identity and lifecycle -- a project, a person, a venture, an experiment, a life goal. Each walnut has its own _kernel/ with identity, history, and knowledge files.
What's the squirrel?
The squirrel is the agent runtime -- the behavioral layer that wraps Claude (or any AI model). It defines how the agent reads your context, captures decisions, routes information, and saves state. The AI model is replaceable. The squirrel runtime persists.
I'm confused / it's not working.
Join @aliveoperators on Discord. We help everyone who's stuck. Or open an issue on GitHub — we respond to everything.