SATO OS — Onchain Agent Mission Control

Build agents on any AI model and any chain. Deploy agents with their own wallets & track agents you already run — all from your own machine.

SPIN UP → FUND → TALK → TASK → WATCH

_{Self-hosted on macOS & Linux (Windows via WSL) · your keys never leave your machine · one-time license, instant download + key}

_{Mission Control at localhost:3300 — status of every agent, wallet, policy, and pending decision in this workspace.}

---

The loop: five verbs, real funds

01 · Spin up	Name it, give it a job in one sentence, mint it a wallet sealed on your machine.
02 · Fund	Send an allowance — lunch money, never the treasury. Spend caps are mandatory.
03 · Talk	Chat or voice. It knows its balance, its policy, and exactly what it's allowed to do.
04 · Task	Give it a schedule and a budget envelope, then let it run while you sleep.
05 · Watch	Every action is re-checked against the chain before it earns a verified receipt.

Bring your own everything

• ⬡ Any AI model — Anthropic, OpenAI, OpenRouter, Nous, xAI. One brain per agent.
• ◇ Any chain — Ethereum, Base, Solana, Arbitrum, Optimism, Polygon, live out of the box. Transfers, swaps, x402 micropayments, ERC-8004 identity, and scheduled tasks wired in on Ethereum & Base.
• ✦ Track agents you already run — attach any agent by its wallet address. No migration.

_{Every agent in the workspace: mode, status, chains, wallet posture, and the brain it runs on.}

All the onchain-agent standards, out of the box

Wired into every agent from birth — live on Base today, not a roadmap slide.

• ERC-8004 — identity at birth. Every agent is born registered on-chain, with its own agentId and a machine-readable agent card.
• EIP-7702 — smart accounts. Approve and swap settle as one atomic transaction.
• ERC-7715 — session grants. Sign once; the agent gets a scoped, capped, expiring permission the chain itself enforces. Over-cap spends are refused by the chain, and revocation always wins.
• EAS — a track record. Verified executions become on-chain attestations — a reputation your Agent Passport can read.

Every action stops at your approval

Every onchain action an agent wants to take stops here first — or, on signer wallets, executes inside your policy tiers with a veto window. Veto and freeze always win.

_{The approvals queue — each proposal carries its risk read and policy verdict. Note the FAIL rows: the policy engine refusing over-cap and post-revocation spends.}

PROPOSE → POLICY CHECK → EXECUTE → CHAIN CHECK → ✓ VERIFIED

External agents are propose-only. No execute tool exists at any policy tier, by construction. Every settled transaction is checked back against the proposal — sender, target, amount — before it earns a verified receipt.

Hermes — fleet command

One sentence. A whole fleet dispatched. Hermes is mission control's voice: it turns one ask into an itemized plan across your agents — you approve once, it dispatches every task, monitors the runs, and reports back. Structurally unable to move money — Hermes commands, never spends.

Your machine. Your keys. Sealed.

Sato OS runs locally — no cloud dependency, no custodian. Each agent's key is generated and sealed on your device (Argon2id + AES-GCM). Caps are mandatory: no caps, no signer.

Control	What it does
$ caps	per-transaction & daily limits on every signer
Veto	a time-locked window on every autonomous action
Kill switch	freeze one wallet — or everything — instantly

Memory you can see

Every note your agents remember lives in an inspectable constellation — browse it, trace it, prune it, export the whole graph to Obsidian. Nothing hides in a black box.

Become a Founding Operator

• ✓ Self-hosted — keys never leave your machine
• ✓ Every agent, wallet, policy tier & rail
• ✓ All 1.x updates — instant download + key
• ✓ Runs on macOS & Linux · Windows via WSL
• ✕ No monthly subscription

Get your license → satohub.ai/os

_{Pay with card or crypto (USDC) · secure checkout by Stripe · add a wallet for your Founding-Operator on-chain credential}

Support

• 🐛 Bugs & questions → Issues
• 🔒 Security → SECURITY (security@satohub.ai)
• 📦 Release notes → Releases

This repository is Sato OS's public home — release notes, docs, and support. Sato OS is commercial, self-hosted software; the source is not published here.

Where it sits in the index

Sato OS is listed in satohubai/onchain-agents and on satohub.ai and scored by the same Sato Score rubric as every other entry — no home-team bonus. That's the point of the score.

---

_{⚠️ Experimental software that signs real onchain transactions under caps you set — start small. Screenshots are the real product running on seeded demo data; figures are wallet-observed demo values, not performance claims. Nothing here is financial advice, and no agent is presented as profitable or safe. © 2026 Sato Hub · proprietary, source-available. Built by Sato Hub — the builder, library & Agent Passport behind the OS.}