Configuration hub for AI coding tools

One directory.
Every AI tool.

jato keeps your instructions, skills, MCPs, and agents in one place. Switch context with a command. Build your setup conversationally with your AI.

$ npx jato init
The superpower

Your AI builds your jato
through conversation.

jato installs a skill that teaches your LLM how to create, modify, and manage your entire setup. Describe what you need — it writes the configs, generates the skills, picks the right MCPs.

r
jato manager
skill active
Architecture

Two skills. Two jobs.
One seamless experience.

Global skill
jato manager

Installed once. Teaches the LLM to be your configuration wizard.

◆ Create new jatos conversationally
◆ Add MCPs, skills, agents on demand
◆ Switch between jatos
◆ Run diagnostics and health checks
◆ Knows 10+ common MCP configurations
Per-jato skill
jato context

Generated fresh on every switch. Gives the LLM your current setup context.

◆ Active jato name and description
◆ Available skills with file paths
◆ Configured MCPs and their env vars
◆ Global and provider-specific instructions
◆ Agent definitions
💡 The jato manager skill is the real product. The CLI is plumbing — the LLM is the interface.
How it works

A jato is a folder.
Everything inside is your AI setup.

▸ ~/.jato/
● config.yaml
▸ skills/
● jato-manager.md
▸ rigs/
▸ mobile/
● jato.yaml
● instructions.md
▸ providers/
● claude.md
● codex.md
▸ skills/
● code-review.md
● testing-rn.md
● .env.example
▸ backend/
…
jato.yaml

The manifest declares which providers are active, what MCPs to configure, and permissions. Minimal by design.

providers/

Provider-specific instruction files. claude.md becomes your CLAUDE.md. codex.md becomes AGENTS.md.

skills/

Markdown skills files. Discovered automatically. Available to every LLM through the installed context skill.

One command

Switch your entire AI environment
in the time it takes to type it.

Terminal
$ jato use mobile
✓ Profile loaded — mobile ✓ Claude Code → ~/.claude/settings.json ✓ Claude Code → CLAUDE.md ✓ Codex CLI → ~/.codex/config.toml ✓ Codex CLI → AGENTS.md ✓ Skills installed — jato-context, jato-manager Active jato: mobile
Backup
Every file gets a timestamped backup before writes. Always.
Atomic
Writes go to .tmp first, then rename. No half-written configs.
Idempotent
Run it twice, get the same result. No state accumulation.
Getting started

Three ways to start.
Zero starts from scratch.

Import
jato init --from auto

Scans your Claude, Codex, Gemini configs. Extracts MCPs, instructions, and skills into a jato.

Template
jato init --template mobile

Pick a starter jato: mobile, backend, fullstack. Pre-configured MCPs, skills, and instructions.

Wizard
just ask your AI

Once jato is installed, ask your LLM to build a jato from your description. It writes everything.

Providers

jato speaks every format
so you don't have to.

Claude Code
→ settings.json → CLAUDE.md → skills/jato-context → skills/jato-manager
Codex CLI
→ config.toml → AGENTS.md → skills/jato-context → skills/jato-manager
Gemini CLI
→ settings.json → GEMINI.md
OpenCode
→ opencode.json
Teams

A jato is a folder.
Share it like you share code.

New team member
$ jato install github.com/acme/ai-jato --jato mobile
✓ Cloned acme/ai-jato ✓ Installed jato: mobile
$ jato use mobile
✓ Ready. Same tools, same context, zero config.

No sync services. No cloud accounts. No proprietary formats. A git repo is your distribution mechanism.

CLI

Six commands.
That's the whole surface.

jato init Import, template, or empty — create your first jato
jato use <jato> Activate a jato — materialize configs and install skills
jato list Show available jatos
jato off Deactivate the current jato
jato doctor Health check — schema, env vars, providers, skills
jato install <repo> Install jatos from a git repository

Stop configuring.
Start building.

Open source. MIT licensed. Let your AI do the setup.

View on GitHub
npx jato init