Personal AI should be portable.

Most AI assistants become more useful by accumulating more data about you inside a proprietary system.

We take a different approach.

Your agent’s purpose, context, skills, permissions, decisions, and handoffs can live in simple human-readable files. That makes agents easier to inspect, version, move between environments, specialize, and combine.

Modular by design

Start with one job.

Add skills, tools, models, project context, or independent review only when they are useful.

A personal assistant can remain simple. A project agent can maintain long-running context. Specialist agents can execute defined workflows. Together, they can form a personalized agent fleet.

Privacy by architecture

The system does not need to build a permanent hidden profile in order to personalize an agent.

The important operating context can remain explicit, portable, and user-controlled.

Autonomy without surrendering control

Agents can be given different levels of authority:

Observe → Draft → Propose → Execute

Access, spending, tools, projects, and external actions can all remain bounded by explicit rules and approval gates.

Capability can expand. Authority remains yours.

Portable agent state

A simple agent might look like:

MY_AGENT/
├── AGENT.md
├── CONTEXT.md
├── SKILLS.md
├── DECISIONS.md
└── HANDOFF.md

No proprietary personality blob required.

The goal is simple:

Build AI that becomes more useful because you can shape it — not because it quietly owns more of your data.