javi-dots

Developer workstation setup — AI CLIs, SDD, memory, and code review in one command

Quick Start

npx javi-dots

That's it. An interactive TUI walks you through selecting AI CLIs, then installs everything: skills, configs, orchestrators, persistent memory, SDD workflow, and optionally code review.

What It Does

javi-dots is the single entry point for setting up an AI-powered developer workstation. It orchestrates multiple tools so you don't have to install them one by one.

flowchart LR
    A["npx javi-dots"] --> B["Select AI CLIs"]
    B --> C["javi-ai install"]
    C --> D["agent-teams-lite<br/>(SDD workflow)"]
    D --> E["engram<br/>(persistent memory)"]
    E --> F{"ghagga?"}
    F -- yes --> G["ghagga init<br/>(code review)"]
    F -- no --> H["Done ✓"]
    G --> H

Loading

Setup Steps

Step	Component	Required	Description
1	javi-ai	✅	Installs skills, configs, and orchestrators for selected CLIs
2	agent-teams-lite	✅	Clones and configures the SDD (Spec-Driven Development) framework
3	engram	✅	Installs persistent AI memory via Homebrew, configures per CLI
4	ghagga	❌	Optional multi-agent code review system
5	rtk	❌	Token compressor — intercepts shell output, saves 60-90% context tokens

Presets

Preset	CLIs	ghagga	TUI
full	All 6 (Claude, OpenCode, Gemini, Qwen, Codex, Copilot)	✅	Skipped
minimal	Claude only	❌	Skipped
custom	Interactive selection	Interactive	Full TUI

# Full preset — everything, no prompts
npx javi-dots --preset full

# Minimal — Claude only
npx javi-dots --preset minimal

# Custom — interactive TUI (default)
npx javi-dots

All commands work in non-TTY environments (pipes, subprocesses).

Commands

Command	Description
setup	Set up developer workstation (default)
doctor	Show health report of current installation
update	Re-run setup for previously configured CLIs
uninstall	Remove javi-dots managed files and manifests
health	Audit AI agent config quality (CLAUDE.md, skills, MCP, hooks)
esp	Claude ESP tmux integration (toggle split pane with ESP watcher)
mcp	MCP server auto-setup (bootstraps default servers: engram, filesystem, glance)
tokens	Token tracking with .wolf/ ledger, repeated read detection
nano <desc>	SDD-lite inline workflow (challenge, plan, build, review)
efficiency on <id>	Activate token-efficiency profile (concise, automation, exploratory)
efficiency off	Deactivate current efficiency profile
efficiency list	List available efficiency profiles
efficiency status	Show current efficiency profile

npx javi-dots setup        # same as just: npx javi-dots
npx javi-dots doctor       # check installation health
npx javi-dots update       # re-apply with same config
npx javi-dots uninstall    # clean removal
npx javi-dots health       # audit CLAUDE.md, skills, MCP config, hooks
npx javi-dots esp          # install ESP tmux toggle (Ctrl-e)
npx javi-dots mcp          # bootstrap default MCP servers
npx javi-dots tokens       # show token usage report from .wolf/ ledger
npx javi-dots nano "add error boundary"  # SDD-lite: challenge → plan → build → review
npx javi-dots efficiency on concise      # activate concise mode (~63% output reduction)
npx javi-dots efficiency off             # back to normal mode

CLI Flags

Flag	Type	Default	Description
--dry-run	boolean	false	Preview changes without writing anything
--preset	string	custom	Preset: full, minimal, custom
--cli	string	—	Comma-separated CLIs: claude,opencode,gemini,qwen,codex,copilot
--ghagga	boolean	—	Enable ghagga code review
--no-ghagga	boolean	—	Disable ghagga code review

Flags override presets. If you pass both --cli and --ghagga/--no-ghagga, the TUI is skipped entirely:

npx javi-dots setup --cli claude,opencode --ghagga --dry-run

What Gets Installed

javi-ai

The AI development layer installer. Deploys skills, configs, orchestrators, hooks, and plugins to ~/.claude/, ~/.config/opencode/, ~/.gemini/, etc. depending on which CLIs you selected.

👉 See javi-ai

agent-teams-lite (SDD)

Spec-Driven Development framework. Cloned to ~/.javidots/agent-teams-lite/ and configured for each selected CLI using mapped names (claude → claude-code, gemini → gemini-cli). Provides structured planning workflows: proposal → spec → design → tasks → apply → verify.

engram

Persistent AI memory system. Installed via Homebrew (brew install gentleman-programming/tap/engram). Each CLI gets its own engram configuration so conversations persist across sessions.

rtk (optional)

Rust Token Killer. Intercepts shell command output and compresses it before it reaches the LLM context window, saving 60-90% of tokens on verbose command output. Installed via Homebrew (brew install rtk-ai/tap/rtk) or Cargo (cargo install rtk). During setup, rtk init -g creates a global config.

To toggle RTK on/off per session, use rtk on / rtk off in your shell.

ghagga (optional)

Multi-agent code review system. When enabled, runs ghagga init to configure consensus-based AI code review in your projects.

👉 See ghagga

Manifest

Configuration is saved to ~/.javidots/manifest.json. The update and uninstall commands read this manifest to know what was installed.

Requirements

• Node.js ≥ 18
• git — required for cloning agent-teams-lite
• brew — required for installing engram (macOS/Linux)

Ecosystem

Package	Description
javi-dots	Workstation setup (this package)
javi-ai	AI development layer — skills, configs, orchestrators
javi-forge	Project scaffolding — CI, templates, AI bootstrap

License

MIT