Trust Layer for Autonomous Finance
Settlement moves money.
Authorization grants permission.
ChainSage decides whether it should happen.
The decision engine between an AI agent's intent and on-chain execution. Every transaction is simulated, scored, and checked against policy and a shared trust network — then returned as a single verdict: ALLOW, REVIEW, or DENY.
Swap 2.5 ETH → USDCUniswap v3 · 0.3% poolTwo layers shipped. The third is still missing.
Autonomous finance can already move money and grant agents permission to spend it. What it can't do is decide whether a given action should be allowed.
moves money
Stablecoins, rollups, and payment rails moved value on-chain in seconds. Solved.
grants permission
Account abstraction, session keys, and x402 let agents hold and spend. Solved.
decides if it should happen
Nothing asks the one question that matters before an agent signs: should this transaction happen at all?
One pipeline, from intent to execution
Every agent action walks the same seven steps. The verdict is the gate — execution only happens on the far side of a decision.
- 01Agent Intent
An agent proposes a transaction.
- 02Simulation
Execute it in a fork — see the real outcome before it happens.
- 03Risk Engine
Score approvals, drainers, contract age, and value at risk.
- 04Policy Engine
Check the action against the owner's runtime rules.
- 05Trust Network
Weigh the counterparties' shared reputation.
- 06Verdictthe gate
ALLOW · REVIEW · DENY — one decision, with reasons.
- 07Execution
Only an ALLOW (or confirmed REVIEW) ever reaches the chain.
Meet your ChainSage Agent
A wallet copilot that reads your holdings, scores every move, and answers with a verdict — not just a number. Try it.
Hey — I'm your ChainSage Agent. I read this wallet live and give you a verdict before you act. Ask me anything, or tap a suggestion.
Quick risk check on your wallet:
Illustrative demo · canned responses · not connected to a live wallet
From a consumer guardian to a category standard
Five phases, each a route you can open now. Guardian is live on Base; the SDK and Risk API ship as developer docs; the Policy Engine and Trust Network are explorable previews running on real, tested engines.
A shared memory for the agent economy
Every verdict feeds a reputation graph. When one node turns malicious, the whole network knows in seconds — and every connected wallet is protected at once.
Idle — run the simulation to stream live signals.
The rails are here. The judgment isn't.
Agents can now hold money and spend it autonomously. Every new protocol makes execution faster — and the missing decision layer more urgent.
HTTP-native payments let agents pay per request. Money now moves at machine speed.
Agent Payments Protocol standardizes how agents authorize and settle on a user's behalf.
Autonomous agents are gaining wallets, mandates, and the ability to act without a human in the loop.
The judgment layer those rails are missing — the verdict between an agent's intent and the chain.
Execution, meet judgment.
Agents on Virtuals execute. ChainSage decides whether they should — an ACP verdict service other agents call before they act.
$SAGElaunches on Virtuals via Genesis. We're building ChainSage as an ACP verdict service — integration in progress, not yet live.
Help Build the Trust Layer for Autonomous Finance.
Guardian is live on Base today — scan any wallet, read-only, and get a verdict on its approval surface. The rest of the trust layer is being built in the open.
Read-only · keys never touched · no funds ever moved