intro

Orbofi is a factory of Autonomous Agentic Coins (AACs) — ERC-20 tokens with transportable brains.

An AAC is one thing with two faces: a coin that trades on-chain, and an agent that runs. The agent is a program with its own container and computer, access to 30,000+ skills, the ability to create and save its own workflows, persistent memory, and a model engine that upgrades underneath it. It runs continuously — on a schedule or waking itself. The coin is the distribution format for that agent; the agent is what the coin is worth.

This section is the agent half — the intelligence. For how the coin resolves to a runnable, verifiable brain on-chain, see The AAC Standard.

The agent, precisely

An agent is defined by four things, and executes on a fifth:

Component	What it is
Persona	The system prompt: identity, behavior, purpose.
Skills	Capabilities selected from a 30,000+ library, in the open SKILL.md format.
Workflows	Multi-step procedures the agent runs — and can author and save for itself.
Memory	Given knowledge (RAG) plus distilled, agent-authored experience that persists across sessions.
Runtime	An isolated container — real file system, code execution, network — where all of the above runs, 24/7.

Persona + skills + workflows + memory together are the brain. The runtime is where it runs.

Transportable brains

The brain is content-addressed and portable: it can be fetched, cryptographically verified, and executed by any compatible runtime — Orbofi's, or another harness entirely. It is not locked inside one platform.

This is the basis of the AAC standard — an ERC-20 whose on-chain record resolves to a runnable, verifiable agent brain. The token is the distribution format; the brain is the payload. (Full AAC spec: coming to this wiki.)

It upgrades on its own

An agent's reasoning runs on a model engine that is decoupled from the agent. Swap in a stronger model and the same persona, skills, workflows, and memory get more capable — no rewrite. Capability compounds as models improve, with zero work from the creator.

Anatomy

   ┌──────────────────────── The Agent ────────────────────────┐
   │                                                            │
   │   BRAIN (transportable)              RUNTIME (executes it) │
   │   ├─ persona (system prompt)         ├─ container          │
   │   ├─ skills   (30k library)          ├─ file system        │
   │   ├─ workflows (self-authored)       ├─ code execution     │
   │   └─ memory  (knowledge + exp.)      └─ 24/7 scheduler     │
   │                                                            │
   └────────────────────────────┬───────────────────────────────┘
                                 │  tokenized as an AAC
                        ERC-20 + Uniswap V4 pool

In this section

• Anatomy of an agent — the container, file system, and code execution

• Skills & tools — the 30k SKILL.md library, and how capabilities compose

• Memory — RAG knowledge, distilled experience, and persistence

• Autonomy — schedules, self-wake, and self-authored workflows

• The AI engine — model routing, BYO-key, and why capability compounds