New:Microsoft Teams Notifications Are Now Available in Socket.Learn more →
Get Started

null0

Package Overview
Dependencies
Maintainers
1
Versions
1
Alerts
File Explorer

Advanced tools

Socket logo

Install Socket

Detect and block malicious and high-risk dependencies

Install

null0

Your AI clone - personal context for AI agents

latest
Source
npmnpm
Version
0.1.1
Version published
Maintainers
1
Created
Source

null0

Your AI clone — personal context that travels with you.

null0 is a CLI tool that builds a persistent memory of who you are and injects it into any AI agent you work with. It learns your persona, skills, preferences, and decision-making patterns, then makes that context available everywhere: in chat, in code agents, via clipboard, or through MCP.

Quick Start

Install via curl:

curl -fsSL https://null0.ai/install.sh | bash

Or install via npm:

npm install -g null0

Or install via Homebrew (macOS):

brew tap null0-ai/null0
brew install null0

Then set up and use:

null0 init          # Set up your profile (takes ~2 minutes)
null0               # Launch chat (auto-runs null0 chat if initialized)
null0 remember "I prefer TypeScript over JavaScript"  # Teach null0 something new
null0 visualize     # Open an interactive mind map of your context

Enable verbose error logging:

NULL0_DEBUG=true null0 chat

What null0 Does

FeatureDescription
Personal ContextStores your persona, skills, preferences, hobbies, tone, and memories in structured markdown files
Smart MemoryAuto-categorizes what you tell it into the right context file using AI
Knowledge GraphBacks your memory with Cognee — a semantic knowledge graph that improves search over time
Tone AnalysisLearns your writing style so AI responses match how you actually communicate
Decision SupportGives personalized advice based on your values and past decisions
Clone ModeMakes the AI respond as you, using your full context
Mind Map/visualize generates a draggable, interactive graph of everything the AI knows about you
MCPExposes your context as a Model Context Protocol server
Multi-ProviderWorks with Claude, Codex, and Gemini out of the box

Commands

CLI

null0 init              Interactive setup wizard
null0 chat              Start the chat REPL
null0 remember <text>   Add to memory (auto-categorized)
null0 forget <text>     Remove from memory (fuzzy match)
null0 visualize         Generate and open your context mind map
null0 inject            Output context for pasting into AI sessions
null0 inject -c         Copy context directly to clipboard
null0 status            Show current config
null0 security          Audit file permissions (--fix to repair)
null0 logout            Remove credentials
null0 uninstall         Completely remove null0 and all data

Chat Commands

/help                   Show available commands
/remember <text>        Save to memory (--project for project scope)
/forget <text>          Remove from memory
/train                  Analyze and learn your writing style
/decision <question>    Get advice based on your values
/safeguards <text>      Set boundaries the AI respects
/clone                  Enter AI clone mode
/exit                   Exit clone mode
/projects               Manage projects
/model <name>           Switch AI model
/theme <name>           Change terminal theme
/chats                  Browse saved conversations

How Context Works

null0 stores everything in ~/.null0/:

~/.null0/
├── persona.md          Who you are
├── skills.md           What you know
├── preferences.md      How you like things done
├── hobbies.md          What interests you
├── tone.md             Your writing style (generated by /train)
├── safeguards.md       Boundaries and values
├── decisions.md        Decision-making patterns
├── knowledge/
│   └── global.md       General memories
├── mindmap.json        Last generated mind map graph
├── mindmap.html        Interactive mind map viewer
├── config.json         Provider and model config
└── .cognee/            Cognee knowledge graph data

When you add a memory, null0 uses AI to categorize it into the right file. The Cognee knowledge graph runs in the background, building semantic relationships between everything you've taught it.

The Mind Map

null0 visualize reads all your context, generates a node graph using AI, and opens it in the browser:

  • Drag nodes to rearrange
  • Scroll to zoom in and out
  • Click and drag the background to pan
  • Hover any node for details
  • Color-coded: purple = you, blue = categories, gray = details, green = cross-cutting themes

The graph is only regenerated when you explicitly run visualize — not on every memory update — so it stays fast.

Cognee & Memory

null0 uses Cognee as a semantic knowledge graph layer on top of your context files:

  • cognify — Builds a knowledge graph from your stored text
  • memify — Adds memory algorithms to improve recall
  • search — Finds relevant memories semantically when building context for a prompt

The bridge runs as a Python subprocess with Gemini embeddings (768-dim). It updates in the background after new memories are added, debounced at 30 seconds.

Authentication

null0 stores credentials in your system keychain — nothing is written to disk in plaintext.

ProviderMethod
ClaudeOAuth via claude CLI (setup-token)
CodexBrowser-based OAuth via codex CLI
GeminiAPI key (entered during init)

Architecture

src/
├── auth/               Credential storage and OAuth flows
├── chat/               Streaming chat provider (Claude, Codex, Gemini)
├── commands/           CLI command implementations
│   ├── chat.ts         Chat REPL entry
│   ├── init.ts         Setup wizard
│   ├── remember.ts     Memory addition
│   ├── forget.ts       Memory removal
│   ├── visualize.ts    Mind map generation and rendering
│   ├── code-orchestrator.ts  Iterative code execution
│   └── ...
├── memory/             Context management
│   ├── memory.ts       Add/remove/search memories
│   ├── cognee.ts       Cognee Python bridge
│   ├── ai-context.ts   AI-powered categorization and optimization
│   ├── inject.ts       Context assembly for prompts
│   └── tone-analyzer.ts  Writing style analysis
├── storage/            File I/O, config, audit logging
├── ui/                 Ink-based terminal UI components
├── mcp/                Model Context Protocol server
└── index.ts            CLI entry point (Commander)

Stack: TypeScript, ESM, Commander.js, Ink (React for terminal), Vitest, tsup.

Development

git clone https://github.com/null0-ai/null0-cli.git
cd null0-cli
npm install
npm run build           # Build
npm run dev             # Watch mode
npm test                # Run tests
npm link                # Make `null0` available locally

Tests live in src/**/__tests__/*.test.ts. Use Vitest conventions. Always use .js extensions in imports (NodeNext resolution).

Contributing

Contributions are welcome. Read CONTRIBUTING.md for the full guide. The short version:

  • Open an issue or discussion first
  • Fork, branch, implement, test
  • Commit with Conventional Commits (feat:, fix:, docs:, etc.)
  • Submit a PR

For new commands: add to src/commands/, export from src/commands/index.ts, register in src/index.ts.

Security

  • All credentials stored in system keychain via keytar
  • Context files created with 0600 permissions
  • Run null0 security to audit, null0 security --fix to repair
  • No background processes persist after exit
  • No PII in the open source code

License

MIT — see LICENSE.

Keywords

ai

FAQs

Package last updated on 05 Feb 2026

Related posts